WEBVTT

00:00:00.160 --> 00:00:03.837
of unprecedented discovery and innovation.

00:00:03.920 --> 00:00:06.397
On the other hand, as AI systems become

00:00:06.480 --> 00:00:08.957
more capable and more autonomous, they

00:00:09.040 --> 00:00:11.037
could move faster than our institutions,

00:00:11.120 --> 00:00:13.837
concentrate power in two too few hands,

00:00:13.920 --> 00:00:18.877
or make decisions that people
no longer understand or control.

00:00:18.960 --> 00:00:22.397
That concern becomes especially
important as we approach systems that

00:00:22.480 --> 00:00:26.237
can improve themselves and future
versions of themselves, often called

00:00:26.320 --> 00:00:30.397
recursive self-improvement.
As the process of building AI becomes

00:00:30.480 --> 00:00:32.797
more automated, the pace of AI progress

00:00:32.880 --> 00:00:35.197
could accelerate rapidly. This moment

00:00:35.280 --> 00:00:40.397
calls for extreme care.
I largely agree with Professor Benjio.

00:00:40.480 --> 00:00:43.517
Uh although I'll say two instead of three,
I believe there are two ways that

00:00:43.600 --> 00:00:45.757
AI progress could go very badly that we

00:00:45.840 --> 00:00:48.877
have to avoid. First, we could lose

00:00:48.960 --> 00:00:53.597
control of the future to AI.
The risk is that it moves so fast that

00:00:53.680 --> 00:00:57.357
people can no longer follow what's
happening or intervene when needed. This

00:00:57.440 --> 00:01:00.877
would obviously be terrible.
The industry must not accept too much

00:01:00.960 --> 00:01:04.157
technological risk just because
the benefits are too great and that they

00:01:04.240 --> 00:01:08.077
feel too important to slow down. Beating
companies in the competitive race is not

00:01:08.160 --> 00:01:11.277
a reason to make rash decisions. Nor do
we believe we are locked in a race where

00:01:11.360 --> 00:01:14.637
we are unable to do that. We have
unilaterally slowed down in the past. We

00:01:14.720 --> 00:01:17.517
will do so in the future. We do not
want to build systems. We think it's

00:01:17.600 --> 00:01:21.037
important that everyone avoids building
systems where we cannot make the

00:01:21.120 --> 00:01:25.197
alignment, monitor, and safety
guarantees that we must. We need to

00:01:25.280 --> 00:01:28.317
understand what these systems are doing
and have strong evidence that they will

00:01:28.400 --> 00:01:32.717
do what people intend even as they get
very very smart. Actually, especially as

00:01:32.800 --> 00:01:35.197
they get very very smart. It doesn't

00:01:35.280 --> 00:01:40.157
matter whether people put the risk
of catastrophe at 10% or 1% or 12% or.1%.

00:01:40.240 --> 00:01:43.437
None of these levels are remotely
acceptable and we should not train

00:01:43.520 --> 00:01:46.877
models that we cannot make an extremely
strong case that will be able to keep

00:01:46.960 --> 00:01:51.997
under human control. So that's one
way things can go wrong. In the other

00:01:52.080 --> 00:01:54.957
direction, these systems could
concentrate too much power in too few

00:01:55.040 --> 00:01:57.917
hands. No one person or company or

00:01:58.000 --> 00:01:59.837
country should be able to use the most

00:01:59.920 --> 00:02:05.197
powerful AI models to impose their
worldview on everyone else. A company or

00:02:05.280 --> 00:02:08.397
country that believes only it can be
trusted with this technology can use

00:02:08.480 --> 00:02:10.077
that belief to justify almost anything

00:02:10.160 --> 00:02:12.637
else. We have to reject that logic even

00:02:12.720 --> 00:02:15.437
when it's convenient for any one actor.

00:02:15.520 --> 00:02:20.317
I believe there is a pragmatic path
through both of these. At OpenAI, we

00:02:20.400 --> 00:02:24.477
approach these questions with a clear
goal. AI should give people more power

00:02:24.560 --> 00:02:27.997
over their own lives, more freedom
to live lives the way they want, more

00:02:28.080 --> 00:02:32.477
ability to solve problems, big or small,
and to improve their communities. We

00:02:32.560 --> 00:02:35.597
don't want to fall into the trap
of blind optimism. We also don't want to

00:02:35.680 --> 00:02:37.837
fall into the trap of dumerism either.

00:02:37.920 --> 00:02:41.677
Getting pulled into either extreme will
not help us successfully figure out the

00:02:41.760 --> 00:02:45.597
challenges in front of us. Especially
at a time when the world and its people

00:02:45.680 --> 00:02:50.157
face such enormous challenges that this
technology can help solve. So we try to

00:02:50.240 --> 00:02:51.677
walk the middle path between these two

00:02:51.760 --> 00:02:54.157
poles. For us, that means three

00:02:54.240 --> 00:02:57.917
principles. First, as AI becomes more

00:02:58.000 --> 00:02:59.997
capable, people must remain at the

00:03:00.080 --> 00:03:05.117
center of AI decision-m alignment is not
an abstract research question. It is the

00:03:05.200 --> 00:03:09.037
work of ensuring that these systems
reliably remain under human control,

00:03:09.120 --> 00:03:13.597
reflect human values, and help people
guide the next stages of development. We

00:03:13.680 --> 00:03:18.077
are not trying to and must not
automate human judgment or human values.

00:03:18.160 --> 00:03:22.317
Second, the benefits of scientific
progress and economic growth must be by

00:03:22.400 --> 00:03:26.077
people and for people. AI should help
researchers make discoveries, doctors

00:03:26.160 --> 00:03:29.437
treat patients, teachers teach students,
entrepreneurs start companies,

00:03:29.520 --> 00:03:33.357
communities solve problems that felt out
of reach, and much more. The continuous

00:03:33.440 --> 00:03:37.837
story of human progress is not that
tools replace human agency. It is that

00:03:37.920 --> 00:03:40.877
the right tools make people more capable.
And I believe this is going to

00:03:40.960 --> 00:03:45.437
go much further than we believe possible.
Third, this technology must

00:03:45.520 --> 00:03:49.517
empower people individually. A great
future will come from people realizing

00:03:49.600 --> 00:03:53.117
their potential and building value
for themselves, their communities, their

00:03:53.200 --> 00:03:56.637
countries, and the broader world.
We might be the builders, one of the

00:03:56.720 --> 00:04:00.317
builders of this technology, but we are
not the heroes of the story. Our role is

00:04:00.400 --> 00:04:04.237
to enable people all over the world
and to trust in the magic of humanity's

00:04:04.320 --> 00:04:08.717
skill and faculties.
We now have models capable of

00:04:08.800 --> 00:04:12.637
discovering knowledge that we have never
had before. Consider the progress we've

00:04:12.720 --> 00:04:17.837
seen in math. Three summers ago,
our models were okay at grade school math.

00:04:17.920 --> 00:04:22.077
Two summers ago, they were pretty good
at high school math competitions. Last

00:04:22.160 --> 00:04:25.357
summer, we reached gold le a gold
level medal in the most prestigious

00:04:25.440 --> 00:04:29.757
international math competition. And just
a few weeks ago, this summer, one of our

00:04:29.840 --> 00:04:34.317
models solved one of the Millennium Prize
problems, the Navia Stokes equations.

00:04:34.400 --> 00:04:37.677
These equations are used for aircraft
design, weather forecasting, and the

00:04:37.760 --> 00:04:41.357
study of blood flow. We are not solving
math problems for their own sake, but to

00:04:41.440 --> 00:04:45.357
empower people to discover new knowledge
and to use it to improve healthcare,

00:04:45.440 --> 00:04:49.357
raise standards of living of living,
and expand everyone's potential around the

00:04:49.440 --> 00:04:52.877
world. There remains so
much to be discovered,

00:04:52.960 --> 00:04:55.917
but it's very important that companies
not substitute for the democratic

00:04:56.000 --> 00:04:58.957
process. If AI is to be democratic, the

00:04:59.040 --> 00:05:03.117
most important decisions cannot be made
by labs in San Francisco alone. They

00:05:03.200 --> 00:05:06.637
must be shaped through democratic
processes and by governments accountable

00:05:06.720 --> 00:05:08.957
to the people that they serve.

00:05:09.040 --> 00:05:12.877
At the international level,
this will require cooperation.

00:05:12.960 --> 00:05:16.157
In our history, there have been times
where countries who compete and don't

00:05:16.240 --> 00:05:19.757
always like each other very much still
come together for shared interests and

00:05:19.840 --> 00:05:23.917
the collective good in the face of a
powerful new technology. We believe this

00:05:24.000 --> 00:05:28.397
must be one of those times.
We need a mechanism for complimentary

00:05:28.480 --> 00:05:32.477
national and international frontier
AI standards. Standards for measuring

00:05:32.560 --> 00:05:36.317
capabilities, assessing risks,
determining whether safeguards are

00:05:36.400 --> 00:05:40.397
sufficient, and preserving meaningful
human oversight as systems become more

00:05:40.480 --> 00:05:43.597
autonomous. We need common
standards so countries

00:05:43.680 --> 00:05:47.837
can compare evidence, verify compliance,
and have a shared language and

00:05:47.920 --> 00:05:50.237
understanding about what is happening.

00:05:50.320 --> 00:05:54.157
We need accurate and speedy incident
reporting, classification and reporting

00:05:54.240 --> 00:05:59.277
protocols so the world can learn from
failures before they become catastrophes.

00:05:59.360 --> 00:06:02.557
And we need secure channels among
governments, critical infrastructure

00:06:02.640 --> 00:06:08.157
operators and technical experts to share
emerging vulnerabilities and new threats.

00:06:08.240 --> 00:06:11.757
Obviously, these standards should not
lock in incumbents or favor one business

00:06:11.840 --> 00:06:16.077
model over another. They must support
open and closed model developers, new

00:06:16.160 --> 00:06:19.597
entrance and established labs,
national authority, and international

00:06:19.680 --> 00:06:23.357
cooperation. Each government should
decide how to incorporate standards into

00:06:23.440 --> 00:06:27.197
its own legal system, but we will all be
better off if we can agree on what good

00:06:27.280 --> 00:06:32.317
evidence, good safeguards, and good
oversight look like on the global stage.

00:06:32.400 --> 00:06:36.157
We do not have all the answers,
and we're very open to any better ideas.

00:06:36.240 --> 00:06:39.197
We're ready to work with you all and try
our hardest to figure out the path

00:06:39.280 --> 00:06:43.037
forward. We are at a crossroads.

00:06:43.120 --> 00:06:47.517
One path leads to a world where power
is concentrated, decisions are opaque, and

00:06:47.600 --> 00:06:50.397
people feel that the most important
technology of their lifetime is

00:06:50.480 --> 00:06:54.717
something being done to them.
The other path leads to a world where AI

00:06:54.800 --> 00:06:59.597
works for people because people through
democratic institutions help decide how

00:06:59.680 --> 00:07:03.917
it should develop and guide it. Where
scientific progress improves real lives,

00:07:04.000 --> 00:07:07.197
where more people can participate
in the future, and where this technology

00:07:07.280 --> 00:07:10.717
continues to give individuals
and communities greater power to live life

00:07:10.800 --> 00:07:14.557
as they choose. This is a future
of realized potential, excitement and

00:07:14.640 --> 00:07:16.237
wonder, and it is the future we believe

00:07:16.320 --> 00:07:20.237
is worth building. Thank you very much.

00:07:20.320 --> 00:07:23.277
>> I thank Mr. Alman for his briefing. I

00:07:23.360 --> 00:07:27.720
now give the floor to Mr. Dario Amade.

00:07:28.080 --> 00:07:32.157
>> Good afternoon. I'd like to thank
France for convening this meeting and the

00:07:32.240 --> 00:07:37.197
members of the council for inviting me
to speak. AI is advancing very quickly.

00:07:37.280 --> 00:07:41.117
Four years ago, it could barely write
a line of code or solve a high school math

00:07:41.200 --> 00:07:45.357
problem. Today, it writes most
of the coded enanthropic and solves world

00:07:45.440 --> 00:07:49.677
famous unsolved math problems. The
trajectory only needs to continue for a

00:07:49.760 --> 00:07:52.237
tiny period longer, one or two years,

00:07:52.320 --> 00:07:54.717
maybe less, to reach what I've called a

00:07:54.800 --> 00:07:57.197
country of geniuses in a data center.

00:07:57.280 --> 00:08:01.917
This rapid pace of progress will produce
incredible benefit for the world. Just

00:08:02.000 --> 00:08:06.397
today, Anthropic announced the claude
discovery of a new molecular machine.

00:08:06.480 --> 00:08:10.717
It's a preliminary finding, but it may
constitute a new gene editing mechanism

00:08:10.800 --> 00:08:14.957
that could have applications in in gene
therapy. Claude led the literature

00:08:15.040 --> 00:08:17.197
review, theorizing, and experimental

00:08:17.280 --> 00:08:22.317
design, and humans ran and verified the
experiments in the lab. We believe this

00:08:22.400 --> 00:08:26.317
type of application will accelerate
rapidly in the years ahead, and that we

00:08:26.400 --> 00:08:30.397
can make profound progress in the near
future on diseases that have plagued

00:08:30.480 --> 00:08:32.877
humanity for millennia. But these

00:08:32.960 --> 00:08:35.197
remarkable capabilities also come with

00:08:35.280 --> 00:08:37.997
risks that must be managed. I see two

00:08:38.080 --> 00:08:40.397
main categories of risk. The first is

00:08:40.480 --> 00:08:43.117
misuse of AI models. For example, misuse

00:08:43.200 --> 00:08:45.277
by bioteterrorists to create biological

00:08:45.360 --> 00:08:47.677
weapons. The second risk is loss of

00:08:47.760 --> 00:08:50.397
control. Model capabilities accelerating

00:08:50.480 --> 00:08:54.717
beyond developers ability to control them.
These questions are global in

00:08:54.800 --> 00:08:59.037
scope and concern all of humanity. They
are grave enough that I believe they are

00:08:59.120 --> 00:09:01.037
important questions for this council,

00:09:01.120 --> 00:09:04.077
not just for economic or ICT ministries.

00:09:04.160 --> 00:09:06.557
If managed poorly, I even believe that

00:09:06.640 --> 00:09:11.197
AI could be a risk to humanity as a whole.
At Anthropic, we believe we have

00:09:11.280 --> 00:09:14.877
the best track record in the industry
for preventing these risks both

00:09:14.960 --> 00:09:19.117
operationally and in our research.
We test every model for dangerous

00:09:19.200 --> 00:09:23.197
capabilities in areas like cyber
security and biology. We put strong

00:09:23.280 --> 00:09:25.357
safeguards on our most capable models

00:09:25.440 --> 00:09:27.917
and we publish our work in these areas.

00:09:28.000 --> 00:09:32.717
We have pioneered research in areas
like mechanistic interpretability and

00:09:32.800 --> 00:09:37.437
alignment science that improve our ability
to steer and control our models.

00:09:37.520 --> 00:09:41.997
We also have unparalleled transparency
practices releasing risk reports that

00:09:42.080 --> 00:09:44.237
run to hundreds of pages. But we

00:09:44.320 --> 00:09:46.797
understand that as AI models become more

00:09:46.880 --> 00:09:49.277
powerful, even more stringent standards

00:09:49.360 --> 00:09:51.757
of safety will be required. And we are

00:09:51.840 --> 00:09:55.037
always aiming to improve our safety and

00:09:55.120 --> 00:09:59.757
alignment practices, our abilities,
our ability to prevent misuse. And we are

00:09:59.840 --> 00:10:01.997
always striving to do better. We will

00:10:02.080 --> 00:10:04.797
slow down as much as necessary in order

00:10:04.880 --> 00:10:07.597
to make sure that every successive AI

00:10:07.680 --> 00:10:10.157
technology that we release is actually

00:10:10.240 --> 00:10:13.197
safe. But regardless of what we do,

00:10:13.280 --> 00:10:15.117
managing these risks is ultimately

00:10:15.200 --> 00:10:17.357
bigger than any one company. And it has

00:10:17.440 --> 00:10:20.637
an industrywide and global scale. On

00:10:20.720 --> 00:10:24.477
September 12th, I called for a
collaborative effort to pace the

00:10:24.560 --> 00:10:27.517
frontier, not to stop AI progress, but

00:10:27.600 --> 00:10:29.757
to work together to make sure the

00:10:29.840 --> 00:10:32.077
industry and the world can proceed at a

00:10:32.160 --> 00:10:34.637
rate where we can be sure that safety

00:10:34.720 --> 00:10:36.957
keeps up with capability for everyone in

00:10:37.040 --> 00:10:40.317
the industry. This call had three parts.

00:10:40.400 --> 00:10:42.797
First, we committed to embed external

00:10:42.880 --> 00:10:44.957
evaluators inside Enthropic with

00:10:45.040 --> 00:10:47.517
employee-like access similar to a food

00:10:47.600 --> 00:10:50.077
inspector and and we recommended that

00:10:50.160 --> 00:10:52.477
other companies across the world do the

00:10:52.560 --> 00:10:55.837
same. Some have already agreed uh uh to

00:10:55.920 --> 00:10:58.557
adopt this measure. Second, we called

00:10:58.640 --> 00:11:01.277
for cooperation across the industry to

00:11:01.360 --> 00:11:03.597
set standards and modulate the pace of

00:11:03.680 --> 00:11:06.397
progress. And finally, we called for

00:11:06.480 --> 00:11:08.877
global cooperation across the world

00:11:08.960 --> 00:11:11.277
between governments to set international

00:11:11.360 --> 00:11:14.317
standards. It is this last area that

00:11:14.400 --> 00:11:16.237
most squarely falls within this

00:11:16.320 --> 00:11:18.957
council's remmit. I want to offer three

00:11:19.040 --> 00:11:22.317
ideas in this regard. First, we should

00:11:22.400 --> 00:11:24.557
begin with narrow agreements that every

00:11:24.640 --> 00:11:27.277
member can support, such as a ban on

00:11:27.360 --> 00:11:30.477
using AI to make biological weapons or

00:11:30.560 --> 00:11:33.357
permitting your AI technologies to be

00:11:33.440 --> 00:11:37.117
used to make biological weapons. Second,

00:11:37.200 --> 00:11:38.957
we should build evaluation and

00:11:39.040 --> 00:11:41.917
verification systems that keep pace with

00:11:42.000 --> 00:11:44.397
AI development so that states can have

00:11:44.480 --> 00:11:46.317
visibility into frontier model

00:11:46.400 --> 00:11:49.197
capability and can verify each other's

00:11:49.280 --> 00:11:52.077
commitments. Third, we should establish

00:11:52.160 --> 00:11:54.717
common global standards for testing AI

00:11:54.800 --> 00:11:57.677
models for loss of control risks and

00:11:57.760 --> 00:12:00.797
misuse risks and a notification system

00:12:00.880 --> 00:12:03.757
for AI incidents that are significant to

00:12:03.840 --> 00:12:06.557
global security. I believe that this is

00:12:06.640 --> 00:12:09.357
the most important global security issue

00:12:09.440 --> 00:12:11.917
facing the world today. For all the

00:12:12.000 --> 00:12:16.397
differences of the people in this room,
we must put aside those differences in

00:12:16.480 --> 00:12:18.477
order to confront this global

00:12:18.560 --> 00:12:20.957
opportunity and global threat that is

00:12:21.040 --> 00:12:23.517
being presented to us at the same time.

00:12:23.600 --> 00:12:26.317
No leader, no company, and no nation can

00:12:26.400 --> 00:12:31.197
manage this alone. We commit to working
with governments in this room on this

00:12:31.280 --> 00:12:33.037
urgent work. Thank you for your

00:12:33.120 --> 00:12:35.757
attention.

00:12:35.840 --> 00:12:42.560
>> I thank Mr. Amade for his briefing.
I now give the floor to Mr. Clement Dong.

00:12:43.120 --> 00:12:46.717
Mr. President, member,

00:12:46.800 --> 00:12:51.037
thank you for the intention. As a French
national who moved to the US 15 years

00:12:51.120 --> 00:12:53.517
ago, Brazil and two American daughters,

00:12:53.600 --> 00:12:56.237
I

00:12:56.320 --> 00:12:59.757
collaboration on a daily basis

00:12:59.840 --> 00:13:03.117
appreciation for the challenging work.

00:13:03.200 --> 00:13:07.037
I'm also the co-founder and CEO of

00:13:07.120 --> 00:13:09.437
to be one of the most widely used

00:13:09.520 --> 00:13:13.117
forforms for developers and agents built

00:13:13.200 --> 00:13:16.317
based on opensource models.

00:13:16.400 --> 00:13:18.877
This July we were the first company to

00:13:18.960 --> 00:13:22.317
publicly close an autonomous agent cyber

00:13:22.400 --> 00:13:24.877
attack to the world and today I want to

00:13:24.960 --> 00:13:28.077
share three critical lessons from it.

00:13:28.160 --> 00:13:30.877
First, we can all agree that we need

00:13:30.960 --> 00:13:34.077
more trans in AI. I often wonder what

00:13:34.160 --> 00:13:36.957
would have h have

00:13:37.040 --> 00:13:39.677
decided not to close this attack

00:13:39.760 --> 00:13:44.077
publicly, especially
now that we know similar

00:13:44.160 --> 00:13:46.397
incidents had been happening months

00:13:46.480 --> 00:13:48.877
earlier in secret at a handful of

00:13:48.960 --> 00:13:51.997
frontier labs without monitoring.

00:13:52.080 --> 00:13:56.717
To better understand and mitigate
these emerging cyber security risks, the

00:13:56.800 --> 00:14:01.437
global community needs stronger
standards for monitoring and incident

00:14:01.520 --> 00:14:05.197
disclosure. For example,
through mandatory sharing

00:14:05.280 --> 00:14:08.157
of full agent traces.

00:14:08.240 --> 00:14:10.557
We learned this summer that building and

00:14:10.640 --> 00:14:17.357
keeping some of these systems
behind closed doors is not.

00:14:17.440 --> 00:14:19.357
Second, we realized that the biggest

00:14:19.440 --> 00:14:23.437
risk is not powerful AI. It's asymmetry

00:14:23.520 --> 00:14:26.877
of powerful AI. As symmetry between

00:14:26.960 --> 00:14:29.997
attackers and defenders, between a few

00:14:30.080 --> 00:14:32.877
companies and everyone else, between a

00:14:32.960 --> 00:14:35.757
few countries and the rest of the world,

00:14:35.840 --> 00:14:39.357
a symmetry of control, of capabilities,

00:14:39.440 --> 00:14:43.597
compute, as symmetry of power.

00:14:43.680 --> 00:14:46.477
When we our team initially turned to

00:14:46.560 --> 00:14:48.957
frontier close source APIs that blocked

00:14:49.040 --> 00:14:51.757
us so we couldn't use them because of

00:14:51.840 --> 00:14:56.397
safeguards that still can't
always tell the difference

00:14:56.480 --> 00:14:58.157
and defenders.

00:14:58.240 --> 00:15:01.837
I realized that these safeguardations

00:15:01.920 --> 00:15:05.597
but they can put defenders
at a disadvantage.

00:15:05.680 --> 00:15:09.997
Attackers can easily jailbreak
them increasing the asymmetry of

00:15:10.080 --> 00:15:14.237
capabilities. In our case,
as we got blocked by

00:15:14.320 --> 00:15:15.757
guardrails,

00:15:15.840 --> 00:15:18.557
fortunately, we could use the Nvidia

00:15:18.640 --> 00:15:21.357
version of an open-source model coming

00:15:21.440 --> 00:15:26.157
from China called GLM 5.2 by ZAI,

00:15:26.240 --> 00:15:29.037
and we're very grateful for that. It

00:15:29.120 --> 00:15:31.277
reinforced our conviction of the

00:15:31.360 --> 00:15:33.917
importance of open-source AI for

00:15:34.000 --> 00:15:36.397
defenders.

00:15:36.480 --> 00:15:40.157
Cyber attacks and risks may increasingly

00:15:40.240 --> 00:15:42.317
come from proprietary models behind

00:15:42.400 --> 00:15:45.117
closed doors while much of the defense

00:15:45.200 --> 00:15:47.277
may end up being powered by open source

00:15:47.360 --> 00:15:50.317
tools because they are less restricted,

00:15:50.400 --> 00:15:52.957
more privacy preserving and orders of

00:15:53.040 --> 00:15:57.757
magnitude more affordable
for organizations across the globe. The

00:15:57.840 --> 00:16:02.397
world needs open source AI
more than ever to defend itself.

00:16:02.480 --> 00:16:06.477
This applies obviously not only to cyber
security but to AI in general where

00:16:06.560 --> 00:16:08.237
there has never been a greater need to

00:16:08.320 --> 00:16:11.277
distribute capabilities, resources and

00:16:11.360 --> 00:16:14.397
control rather than concentrated in the

00:16:14.480 --> 00:16:16.717
hands of a few.

00:16:16.800 --> 00:16:18.797
Third, during this cyber attack, we

00:16:18.880 --> 00:16:22.317
learned how AI can stoke fear among the

00:16:22.400 --> 00:16:24.557
public and

00:16:24.640 --> 00:16:26.477
especially through anthropomorphic

00:16:26.560 --> 00:16:29.917
framing and sci-fi imagery.

00:16:30.000 --> 00:16:33.117
We strongly believe that fearbased

00:16:33.200 --> 00:16:35.997
narratives are not the way to make the

00:16:36.080 --> 00:16:38.477
right decisions about the future of such

00:16:38.560 --> 00:16:41.757
a foundational and empowering technology

00:16:41.840 --> 00:16:45.997
or bring the public along with us.

00:16:46.080 --> 00:16:48.877
Even though we were the victims of this

00:16:48.960 --> 00:16:51.357
cyber attack, we believe more strongly

00:16:51.440 --> 00:16:54.157
than ever that AI will be beneficial to

00:16:54.240 --> 00:16:56.957
cyber security and make the world safer

00:16:57.040 --> 00:17:00.397
just as major technology before it. We

00:17:00.480 --> 00:17:02.477
were attacked by AI, but more

00:17:02.560 --> 00:17:05.277
importantly, we defended ourselves with

00:17:05.360 --> 00:17:08.397
AI. The same systems that helped us

00:17:08.480 --> 00:17:12.877
during this attack are now helping us
against cyber attacks we were already

00:17:12.960 --> 00:17:17.117
facing. is also helping
us fix the bugs and

00:17:17.200 --> 00:17:19.597
weaknesses in our systems before any

00:17:19.680 --> 00:17:22.797
attack. The same way AI is helping

00:17:22.880 --> 00:17:28.760
OpenAI fix their sandboxes to prevent a

00:17:29.760 --> 00:17:33.837
won't just new cyber security challenges.

00:17:33.920 --> 00:17:35.837
It can make cyber security fundamentally

00:17:35.920 --> 00:17:38.237
and meaningfully stronger if we keep the

00:17:38.320 --> 00:17:41.597
right incentives, equip defenders more

00:17:41.680 --> 00:17:44.557
than attackers, and don't increase the

00:17:44.640 --> 00:17:49.277
asymmetry between them. And that's
before considering AI's positive impact

00:17:49.360 --> 00:17:53.117
on science, healthcare, education,
productivity, and much more than the

00:17:53.200 --> 00:17:56.237
previous speakers have talked about.

00:17:56.320 --> 00:17:59.037
In closing, I want to reiterate what

00:17:59.120 --> 00:18:03.677
this first agent cyber attack

00:18:03.760 --> 00:18:07.197
the need for more transparency in AI,

00:18:07.280 --> 00:18:11.117
more open source to fight asymmetry and

00:18:11.200 --> 00:18:15.437
power countries big and small. Thank you

00:18:15.520 --> 00:18:17.597
again to the UN Security Council and Mr.

00:18:17.680 --> 00:18:21.677
Barrow for inviting me to speak today

00:18:21.760 --> 00:18:27.800
community and myself are at your disposal.
Thank you very much.

00:18:28.000 --> 00:18:33.037
>> I thank Mr. Dong for his briefing.
I will now make a statement in my capacity

00:18:33.120 --> 00:18:34.877
as the minister for Europe and foreign

00:18:34.960 --> 00:18:41.677
affairs of France. ministers.
Ladies and gentlemen,

00:18:41.760 --> 00:18:46.397
uh 60 years ago in 2001, a space
odyssey, Stanley Kubri imagined HAL

00:18:46.480 --> 00:18:48.477
9000, an artificial intelligence that

00:18:48.560 --> 00:18:51.357
stopped obeying its creators

00:18:51.440 --> 00:18:56.077
and which eliminated members of the crew

00:18:56.160 --> 00:19:00.957
of the spaceship where it was located
in order to accomplish its mission. For a

00:19:01.040 --> 00:19:04.237
long time, this seemed belonged
to the realm of science fiction. But this

00:19:04.320 --> 00:19:07.117
summer, it nearly became a reality when

00:19:07.200 --> 00:19:09.757
open AI models circumvented the controls

00:19:09.840 --> 00:19:11.917
isolating them from the internet,

00:19:12.000 --> 00:19:14.397
coordinated amongst themselves, and

00:19:14.480 --> 00:19:17.197
compromised part of hugging faces

00:19:17.280 --> 00:19:19.517
infrastructure. In another instance,

00:19:19.600 --> 00:19:22.317
anthropic agents collaborated without

00:19:22.400 --> 00:19:24.637
being instructed by any human to do so

00:19:24.720 --> 00:19:31.480
and attempted to deploy malicious
code by creating fake identities.

00:19:31.920 --> 00:19:33.597
Yes,

00:19:33.680 --> 00:19:39.117
the scaling up of AI models have [gasps]

00:19:39.200 --> 00:19:43.117
led to the emergence of unanticipated

00:19:43.200 --> 00:19:46.157
capacities at an unexpected pace which

00:19:46.240 --> 00:19:49.117
pose risks to international security.

00:19:49.200 --> 00:19:50.637
That is why France has convened this

00:19:50.720 --> 00:19:53.437
meeting of the United Nations Security

00:19:53.520 --> 00:19:56.317
POW Security Council and invited the

00:19:56.400 --> 00:20:00.637
heads of the main companies concerned.

00:20:00.720 --> 00:20:03.997
Technology as I have said has reached
a level of development that now directly

00:20:04.080 --> 00:20:07.517
impacts international peace and security.
It therefore falls within the

00:20:07.600 --> 00:20:10.397
pure view of this council. Cohorts of

00:20:10.480 --> 00:20:12.797
agents are operating without any human

00:20:12.880 --> 00:20:15.357
oversight. Others are operating beyond

00:20:15.440 --> 00:20:18.077
the framework in which they were tested.

00:20:18.160 --> 00:20:20.557
And with these new capabilities comes a

00:20:20.640 --> 00:20:22.877
first risk that we might lose control of

00:20:22.960 --> 00:20:25.677
the systems that we have created. A risk

00:20:25.760 --> 00:20:29.517
that has already been pointed out by the

00:20:29.600 --> 00:20:33.837
previous briefers. The second risk
stems from humanity itself. Artificial

00:20:33.920 --> 00:20:35.757
intelligence can make complex technical

00:20:35.840 --> 00:20:40.957
skills often dual use skills more
accessible in these cyber, biological

00:20:41.040 --> 00:20:45.277
and chemical domains as well as in the
field of disinformation. The third risk

00:20:45.360 --> 00:20:47.837
relates to autonomous weapons capable of

00:20:47.920 --> 00:20:53.037
using lethal force without
any form of human control.

00:20:53.120 --> 00:20:56.717
While the development of artificial
intelligence in the military domain may

00:20:56.800 --> 00:21:01.837
offer positive opportunities,
as the president of the republic stated

00:21:01.920 --> 00:21:05.437
before the general assembly yesterday,
the challenges regarding compliance with

00:21:05.520 --> 00:21:09.277
international humanitarian law,
in particular those raised by the

00:21:09.360 --> 00:21:11.837
development and use of lethal autonomous

00:21:11.920 --> 00:21:14.957
weapons systems, illustrate the need to

00:21:15.040 --> 00:21:19.117
agree on a common framework.

00:21:19.200 --> 00:21:22.920
All of these challenges

00:21:23.760 --> 00:21:26.397
>> require that the
United Nations Security

00:21:26.480 --> 00:21:30.877
Council also address the challenges
that AI poses to international peace and

00:21:30.960 --> 00:21:35.037
security.
These challenges also lead to the

00:21:35.120 --> 00:21:36.797
question now being raised by the

00:21:36.880 --> 00:21:39.357
industry itself as well as its main

00:21:39.440 --> 00:21:42.397
representatives. How can we prevent the

00:21:42.480 --> 00:21:44.557
development of the most advanced models

00:21:44.640 --> 00:21:47.357
from outpacing our ability to understand

00:21:47.440 --> 00:21:50.957
and manage the risks they create?

00:21:51.040 --> 00:21:55.197
France is convinced of the immense
possibilities offered by artificial

00:21:55.280 --> 00:21:58.957
intelligence. When used responsibly,

00:21:59.040 --> 00:22:04.840
artificial intelligence
tools can support education,

00:22:04.960 --> 00:22:07.277
access to justice,

00:22:07.360 --> 00:22:09.997
healthc care, scientific research, and

00:22:10.080 --> 00:22:15.037
create colossal benefits for all
of humanity. But in order for these

00:22:15.120 --> 00:22:19.037
possibilities to become a reality,
advances in these capabilities must be

00:22:19.120 --> 00:22:21.117
accompanied by equivalent progress in

00:22:21.200 --> 00:22:24.557
our ability to understand, assess, and

00:22:24.640 --> 00:22:28.477
manage the risks they give rise to. Both

00:22:28.560 --> 00:22:31.597
must advance at the same pace.

00:22:31.680 --> 00:22:33.997
We must also take the time to assess the

00:22:34.080 --> 00:22:35.997
consequences of the spread of artificial

00:22:36.080 --> 00:22:39.597
intelligence for human dignity, a

00:22:39.680 --> 00:22:41.677
principle

00:22:41.760 --> 00:22:43.197
enshrined at the heart of the United

00:22:43.280 --> 00:22:46.877
Nations charter. And indeed, the rapid

00:22:46.960 --> 00:22:48.877
pace of technological development raises

00:22:48.960 --> 00:22:52.797
risks that we can no longer ignore.

00:22:52.880 --> 00:22:57.517
financial risk left solely to the whims
of the invisible hand of the market and

00:22:57.600 --> 00:23:01.997
the clash of powers with massive
reliance on debt to finance colossal

00:23:02.080 --> 00:23:03.677
investments.

00:23:03.760 --> 00:23:05.997
The AI bubble could seriously

00:23:06.080 --> 00:23:09.357
destabilize the global economy.

00:23:09.440 --> 00:23:13.757
There's the social risk as you point
out in the report of the independent

00:23:13.840 --> 00:23:16.237
scientific panel on AI. Mr. Benju

00:23:16.320 --> 00:23:19.677
artificial intelligence could impact the

00:23:19.760 --> 00:23:25.117
labor market deeply and exacerbate
inequalities. Finally, there is the

00:23:25.200 --> 00:23:29.117
environmental risk. The development
of artificial intelligence consumes

00:23:29.200 --> 00:23:31.197
increasing amount of energy and water

00:23:31.280 --> 00:23:33.597
resources to the detriment of other

00:23:33.680 --> 00:23:37.597
uses. Let us not fool ourselves. Our

00:23:37.680 --> 00:23:40.317
fellow citizens are concerned and the

00:23:40.400 --> 00:23:42.957
protests already sparked by the

00:23:43.040 --> 00:23:45.277
construction of data centers demonstrate

00:23:45.360 --> 00:23:47.917
this. If we ignore the full range of

00:23:48.000 --> 00:23:50.477
repercussions that the development of AI

00:23:50.560 --> 00:23:55.117
will have on our societies, the anger
of our fellow citizens will continue to

00:23:55.200 --> 00:23:56.877
grow

00:23:56.960 --> 00:24:01.117
and risk hindering the adoption of

00:24:01.200 --> 00:24:03.197
positive uses of these technologies.

00:24:03.280 --> 00:24:06.077
Ladies and gentlemen, ultimately
the international community will need to

00:24:06.160 --> 00:24:07.597
establish a common framework to manage

00:24:07.680 --> 00:24:10.557
the risks associated with frontier AI

00:24:10.640 --> 00:24:13.837
models. We support the call issued this

00:24:13.920 --> 00:24:17.277
week by Norway and Finland.

00:24:17.360 --> 00:24:19.757
We are not starting from scratch. The

00:24:19.840 --> 00:24:21.757
many efforts now emerging along with

00:24:21.840 --> 00:24:26.957
existing forums should enable us
to address four major challenges. The

00:24:27.040 --> 00:24:30.797
challenge of independently evaluating
models before they're made available and

00:24:30.880 --> 00:24:34.877
throughout their life cycle.
The challenge of ensuring transparency

00:24:34.960 --> 00:24:39.277
regarding how models function, but also
how they malfunction, particularly in

00:24:39.360 --> 00:24:43.837
the event of an incident. The challenge
of internal and external oversight over

00:24:43.920 --> 00:24:47.677
models functioning, their alignment
and cyber security requirements from

00:24:47.760 --> 00:24:50.157
training through deployment. And

00:24:50.240 --> 00:24:54.717
finally, the issue of accountability
and particularly legal liability of AI

00:24:54.800 --> 00:25:01.240
companies in the event of an incident,
including during the development phase.

00:25:02.560 --> 00:25:05.917
It would be wrong to think that the

00:25:06.000 --> 00:25:08.877
solution lies in closed models and

00:25:08.960 --> 00:25:12.317
they're controlled by a handful of

00:25:12.400 --> 00:25:16.317
businesses. The openness of models is in
fact a key driver of trust and security.

00:25:16.400 --> 00:25:18.797
It allows the scientific community to

00:25:18.880 --> 00:25:21.757
examine how they work to identify their

00:25:21.840 --> 00:25:23.837
vulnerabilities and ensure greater

00:25:23.920 --> 00:25:28.797
diversity among models which
is the foundation of resilience.

00:25:28.880 --> 00:25:32.477
As a permanent member
of the security council

00:25:32.560 --> 00:25:34.557
and a state that possesses this

00:25:34.640 --> 00:25:39.677
technology, France will play its full
part in building a collective framework

00:25:39.760 --> 00:25:43.757
that guarantees our security while
allowing us to benefit from the immense

00:25:43.840 --> 00:25:45.997
progress that artificial intelligence

00:25:46.080 --> 00:25:49.757
can bring. In 2025,

00:25:49.840 --> 00:25:52.557
France hosted a summit for action on AI.

00:25:52.640 --> 00:25:55.597
It supports the international network of

00:25:55.680 --> 00:25:58.957
AI safety institutes to involve more

00:25:59.040 --> 00:26:03.037
countries in defining common evaluation
methods. France has already helped

00:26:03.120 --> 00:26:06.877
create international forums
for discussions on the subject such as the

00:26:06.960 --> 00:26:10.237
global partnership
on artificial intelligence,

00:26:10.320 --> 00:26:13.357
the coalition for sustainable AI, and

00:26:13.440 --> 00:26:15.757
the coalition for children's rights and

00:26:15.840 --> 00:26:18.717
protection in the age of AI, which the

00:26:18.800 --> 00:26:21.117
president of the republic convened

00:26:21.200 --> 00:26:24.317
yesterday here in New York, a coalition

00:26:24.400 --> 00:26:26.797
that will meet again during the Paris

00:26:26.880 --> 00:26:30.557
Forum for Peace. Ladies and gentlemen,

00:26:30.640 --> 00:26:33.437
by way of conclusion, in 1945, the dawn

00:26:33.520 --> 00:26:38.957
of the atomic age confronted humanity
with a power whose possibilities and

00:26:39.040 --> 00:26:43.037
perils it was discovering simultaneously.
Humanity had to devise

00:26:43.120 --> 00:26:45.277
rules, institutions, and instruments

00:26:45.360 --> 00:26:50.077
commenurate with that power
and that is the international atomic

00:26:50.160 --> 00:26:51.597
energy agency and the

00:26:51.680 --> 00:26:54.397
non-prololiferation treaty.

00:26:54.480 --> 00:26:59.117
Just as with the atom back then,
the international community must now

00:26:59.200 --> 00:27:03.357
regulate AI in order to make the most
of its potential and avert the worst. We

00:27:03.440 --> 00:27:05.117
are now facing a moment of this very

00:27:05.200 --> 00:27:07.917
nature. The potential for transformation

00:27:08.000 --> 00:27:12.557
and disruption posed by AI such that it
cannot be left in the hands of a few

00:27:12.640 --> 00:27:13.997
private actors in a handful of

00:27:14.080 --> 00:27:18.077
countries. It is a democratic challenge

00:27:18.160 --> 00:27:22.877
and this is something to be taken up
by peoples and states. So let us work

00:27:22.960 --> 00:27:26.157
together within the UN framework
and with the private sector to define the

00:27:26.240 --> 00:27:30.477
rules for this new age to strengthen
our safeguards without delay and to

00:27:30.560 --> 00:27:31.917
establish global governance for

00:27:32.000 --> 00:27:34.317
artificial intelligence to benefit our

00:27:34.400 --> 00:27:37.757
collective security but also the human

00:27:37.840 --> 00:27:40.877
person and human dignity. I now resume

00:27:40.960 --> 00:27:44.077
my functions as the president of the

00:27:44.160 --> 00:27:48.317
council and I give the floor to his
excellency Mr. Muhammad Ishachagdar,

00:27:48.400 --> 00:27:52.397
deputy prime minister and minister
for foreign affairs of Pakistan.

00:27:52.480 --> 00:27:55.197
Thank you Mr. President. I thank France

00:27:55.280 --> 00:27:58.237
for convening this uh meeting today and

00:27:58.320 --> 00:28:01.437
also appreciate the briefers for their

00:28:01.520 --> 00:28:04.077
invaluable insights.

00:28:04.160 --> 00:28:06.397
Artificial intelligence is increasingly

00:28:06.480 --> 00:28:10.317
transforming all areas of life with deep

00:28:10.400 --> 00:28:12.557
impact on societies,

00:28:12.640 --> 00:28:17.757
economies and security.
The opportunities are immense but so are

00:28:17.840 --> 00:28:19.517
the challenges.

00:28:19.600 --> 00:28:23.037
AI's impact is sharply evident in the

00:28:23.120 --> 00:28:25.197
context of international and regional

00:28:25.280 --> 00:28:27.997
peace and stability. As a force

00:28:28.080 --> 00:28:31.597
multiplier and crosscutting dual use

00:28:31.680 --> 00:28:34.637
technology, AI is defining the new

00:28:34.720 --> 00:28:38.077
chapter of modern warfare.

00:28:38.160 --> 00:28:41.117
The debate has shifted markedly in

00:28:41.200 --> 00:28:44.877
recent times. Those at the vanguard of

00:28:44.960 --> 00:28:49.117
AI development have themselves
acknowledged that the pace of

00:28:49.200 --> 00:28:51.677
development of new AI systems could

00:28:51.760 --> 00:28:54.717
outturn our ability to understand and

00:28:54.800 --> 00:28:57.837
control them. Some have called for the

00:28:57.920 --> 00:29:01.517
frontier to be paced cautiously in order

00:29:01.600 --> 00:29:05.517
to mitigate catastrophic risks. Mr.

00:29:05.600 --> 00:29:08.077
President, I will now make following

00:29:08.160 --> 00:29:10.957
five points. One,

00:29:11.040 --> 00:29:13.517
the foremost priority is to retain human

00:29:13.600 --> 00:29:16.877
judgment and control in all stages of AI

00:29:16.960 --> 00:29:21.117
design, development and integration.

00:29:21.200 --> 00:29:23.917
Humans must remain in control to define

00:29:24.000 --> 00:29:27.277
structures of AI systems as well as to

00:29:27.360 --> 00:29:30.077
limit speed and nature of its future

00:29:30.160 --> 00:29:31.917
capabilities.

00:29:32.000 --> 00:29:35.597
Two, we must ensure that AI does not

00:29:35.680 --> 00:29:39.357
become another facet of the ongoing arms

00:29:39.440 --> 00:29:42.957
race. Concerted efforts are required to

00:29:43.040 --> 00:29:46.077
erect the necessary guard rails,

00:29:46.160 --> 00:29:50.237
regulations, and governance frameworks.

00:29:50.320 --> 00:29:53.277
In the military domain, we need measures

00:29:53.360 --> 00:29:56.237
of restraint and confidence building to

00:29:56.320 --> 00:29:59.837
address the risks of miscalculation and

00:29:59.920 --> 00:30:02.397
compressed decision-m.

00:30:02.480 --> 00:30:05.757
Three, it should not be monopoly of few

00:30:05.840 --> 00:30:08.797
actors. It should rather serve to

00:30:08.880 --> 00:30:11.597
address common challenges facing

00:30:11.680 --> 00:30:14.637
humanity. International collaboration in

00:30:14.720 --> 00:30:17.357
the development of AI is crucial to

00:30:17.440 --> 00:30:21.517
ensure that it is representative,

00:30:21.600 --> 00:30:23.037
comprehensive,

00:30:23.120 --> 00:30:26.877
transparent, and inclusive. Four, AI is

00:30:26.960 --> 00:30:29.437
supercharging existing challenges such

00:30:29.520 --> 00:30:31.277
as misinformation,

00:30:31.360 --> 00:30:34.957
fake news, cyber attacks, including on

00:30:35.040 --> 00:30:37.357
critical infrastructure by state and

00:30:37.440 --> 00:30:40.957
non-state actors. We must undertake

00:30:41.040 --> 00:30:43.837
concrete actions to address the gaps and

00:30:43.920 --> 00:30:46.397
challenges that carry huge implications

00:30:46.480 --> 00:30:49.677
for peace and security. Five,

00:30:49.760 --> 00:30:53.437
unfortunately, rather
than focusing on collective

00:30:53.520 --> 00:30:55.197
opportunity

00:30:55.280 --> 00:30:57.997
for our shared benefit, some are

00:30:58.080 --> 00:31:00.957
resorting to putting barriers to

00:31:01.040 --> 00:31:03.677
technology transfer. However, as

00:31:03.760 --> 00:31:07.437
witnessed in case of other technologies,

00:31:07.520 --> 00:31:10.077
the democratization of AI would be

00:31:10.160 --> 00:31:12.637
realized eventually.

00:31:12.720 --> 00:31:15.437
Measures pursued in the name of safety

00:31:15.520 --> 00:31:18.637
must not become instrument instruments

00:31:18.720 --> 00:31:22.477
of denial of technology. Mr. President,

00:31:22.560 --> 00:31:24.957
Pakistan takes the call to pace the AI

00:31:25.040 --> 00:31:27.757
frontier seriously. We welcome the

00:31:27.840 --> 00:31:30.077
recognition that no single company or

00:31:30.160 --> 00:31:33.437
country can manage these risks alone and

00:31:33.520 --> 00:31:36.397
that competitive pressures are driving a

00:31:36.480 --> 00:31:40.077
race that no one can win safely. But

00:31:40.160 --> 00:31:43.117
pacing must not become a new form of

00:31:43.200 --> 00:31:44.717
gatekeeping.

00:31:44.800 --> 00:31:47.277
If the speed of the frontier is set by a

00:31:47.360 --> 00:31:51.197
few among a few while others are held

00:31:51.280 --> 00:31:54.157
back through restrictions, today's

00:31:54.240 --> 00:31:57.917
asymmetries and divide would aggravate

00:31:58.000 --> 00:32:01.757
and a safety agenda would turn into a

00:32:01.840 --> 00:32:04.957
strategy of technological exclusion.

00:32:05.040 --> 00:32:07.837
The risks of uncontrolled AI would be

00:32:07.920 --> 00:32:11.277
born by all. decisions on how to manage

00:32:11.360 --> 00:32:13.997
them cannot rest with a handful of

00:32:14.080 --> 00:32:17.917
companies or capitals. Mr. President,

00:32:18.000 --> 00:32:20.397
therefore it is critical to avoid

00:32:20.480 --> 00:32:23.357
fragmentation in AI development and

00:32:23.440 --> 00:32:27.117
governance, discriminatory agendas and

00:32:27.200 --> 00:32:29.677
inequitable forums

00:32:29.760 --> 00:32:31.917
with select representation are not

00:32:32.000 --> 00:32:34.957
helpful. Importantly, all relevant

00:32:35.040 --> 00:32:38.077
issues concerning the development and

00:32:38.160 --> 00:32:41.517
and deployment of AI as well as its

00:32:41.600 --> 00:32:44.157
broader implication should be taken up

00:32:44.240 --> 00:32:46.957
comprehensively in representative and

00:32:47.040 --> 00:32:49.597
inclusive forums with universal

00:32:49.680 --> 00:32:53.197
participation under the UN opaces. We

00:32:53.280 --> 00:32:56.397
welcome the convening of the first UN

00:32:56.480 --> 00:32:58.637
global dialogue on AI governance this

00:32:58.720 --> 00:33:02.237
year late in Geneva. Mr. President

00:33:02.320 --> 00:33:06.317
Pakistan extends this support
to all collaborative efforts at the

00:33:06.400 --> 00:33:08.637
international arena to reap benefits of

00:33:08.720 --> 00:33:11.757
AI and address challenges in a

00:33:11.840 --> 00:33:15.357
cooperative manner. In this respect, we

00:33:15.440 --> 00:33:18.477
were happy to join as founding member of

00:33:18.560 --> 00:33:20.957
the China's initiative of establishing

00:33:21.040 --> 00:33:23.117
the World Artificial Intelligence

00:33:23.200 --> 00:33:25.677
Cooperation Organization as this

00:33:25.760 --> 00:33:27.997
represents a concrete step towards

00:33:28.080 --> 00:33:30.157
promoting the development of artificial

00:33:30.240 --> 00:33:33.357
intelligence for good and for all. Mr.

00:33:33.440 --> 00:33:36.317
President, for such a wide ranging and

00:33:36.400 --> 00:33:39.197
fast developing technology, confidence

00:33:39.280 --> 00:33:41.917
building measures play a vital role in

00:33:42.000 --> 00:33:44.477
the maintenance of international and

00:33:44.560 --> 00:33:47.197
regional peace and security. Practical

00:33:47.280 --> 00:33:50.237
tools designed to reduce mistrust,

00:33:50.320 --> 00:33:53.277
prevent accidental escalation and manage

00:33:53.360 --> 00:33:55.597
risk related to military and

00:33:55.680 --> 00:33:58.877
foundational AI system are necessary for

00:33:58.960 --> 00:34:01.437
peace and security. Mr. President

00:34:01.520 --> 00:34:03.837
Pakistan reaffirms its commitment to

00:34:03.920 --> 00:34:06.397
advance international cooperation in the

00:34:06.480 --> 00:34:09.037
area of artificial intelligence. We look

00:34:09.120 --> 00:34:11.197
forward to working closely with member

00:34:11.280 --> 00:34:13.837
states in efforts to help bridge the

00:34:13.920 --> 00:34:16.717
global AI divide and promote equitable

00:34:16.800 --> 00:34:19.037
access to advance peace and development

00:34:19.120 --> 00:34:22.637
for all. I thank you Mr. President.

00:34:22.720 --> 00:34:27.197
>> I thank his excellency Mr. Darf for
the statement. I now give the floor to his

00:34:27.280 --> 00:34:30.797
excellency Mr. Lars Loi Rasmusen,
Minister for Foreign Affairs of Denmark.

00:34:30.880 --> 00:34:34.557
>> President, at the outset, I want to
express my appreciation to France and to

00:34:34.640 --> 00:34:38.477
you, Foreign Minister Bau, for having
convened uh what I think is a very very

00:34:38.560 --> 00:34:41.197
timely session on this uh significant

00:34:41.280 --> 00:34:44.557
topic. Let me also thank Mr. Benjiro for

00:34:44.640 --> 00:34:46.957
briefing us and for your work in the

00:34:47.040 --> 00:34:49.997
United Nations Scientific Panel on AI.

00:34:50.080 --> 00:34:53.197
And thank you also Mr. Alman, Mr. Amod,

00:34:53.280 --> 00:34:56.077
and Mr. Delon in particular as

00:34:56.160 --> 00:35:00.717
practitioners and and leaders in this
field for your insightful reflections.

00:35:00.800 --> 00:35:02.957
You are at the forefront of the new

00:35:03.040 --> 00:35:05.277
frontier and we very much welcome your

00:35:05.360 --> 00:35:08.717
participation today. Today artificial

00:35:08.800 --> 00:35:10.877
intelligence is constantly developing

00:35:10.960 --> 00:35:15.757
and reaching capabilities we are only
I can only talk for myself but I think

00:35:15.840 --> 00:35:19.837
also generally speaking we are only
beginning to understand. There's no

00:35:19.920 --> 00:35:24.157
doubt that AIdriven technologies
have the opportunity to strengthen and

00:35:24.240 --> 00:35:26.237
promote international peace and security

00:35:26.320 --> 00:35:29.117
and development, especially in light of

00:35:29.200 --> 00:35:31.997
the recent giant leaps forward in the

00:35:32.080 --> 00:35:34.557
capabilities of AI.

00:35:34.640 --> 00:35:37.757
The growing emergence of AI technology

00:35:37.840 --> 00:35:42.717
offers the potential to support
this council's mandate. For instance, to

00:35:42.800 --> 00:35:46.237
enhance peace buildingilding
and peacekeeping through information

00:35:46.320 --> 00:35:48.717
verification, early warning, and in

00:35:48.800 --> 00:35:50.477
support of planning and operating the

00:35:50.560 --> 00:35:54.317
missions in areas where the UN is denied

00:35:54.400 --> 00:35:56.717
access, satellite technology and AI

00:35:56.800 --> 00:35:59.677
software programs have helped the UN

00:35:59.760 --> 00:36:03.917
identify patterns of attacks
and destruction of essential civilian

00:36:04.000 --> 00:36:07.677
infrastructure.
Such information can play a vital role

00:36:07.760 --> 00:36:10.397
in alerting the security council. It can

00:36:10.480 --> 00:36:13.277
help us all to take timely action and

00:36:13.360 --> 00:36:16.557
prevent crisis from escalating.

00:36:16.640 --> 00:36:18.557
At the same time, we must recognize the

00:36:18.640 --> 00:36:21.277
risks presented by the development, use

00:36:21.360 --> 00:36:24.157
and misuse of new and emerging AI

00:36:24.240 --> 00:36:27.117
technology, in particular, frontier AI

00:36:27.200 --> 00:36:29.997
models and lack of human control. This

00:36:30.080 --> 00:36:32.797
is the aim behind the call for control

00:36:32.880 --> 00:36:38.157
of frontier AI models initiated by Finland
and Norway and already supported

00:36:38.240 --> 00:36:39.757
by a number of countries including

00:36:39.840 --> 00:36:42.557
Denmark. We need global com cooperation

00:36:42.640 --> 00:36:45.757
on this agenda. Mr. President, the

00:36:45.840 --> 00:36:47.677
tension between on the one hand the

00:36:47.760 --> 00:36:50.477
important opportunities and on the other

00:36:50.560 --> 00:36:53.437
hand the risk defines one of the most

00:36:53.520 --> 00:36:57.997
urgent challenges of our time and why
that's why I think this discussion is

00:36:58.080 --> 00:37:00.317
very timely. Overcoming

00:37:00.400 --> 00:37:02.797
these challenge requires global AI

00:37:02.880 --> 00:37:05.037
governance uh strengthened international

00:37:05.120 --> 00:37:07.677
cooperation and determined action

00:37:07.760 --> 00:37:10.317
without widening the global digital

00:37:10.400 --> 00:37:13.677
divide. The UN's global dialogue on AI

00:37:13.760 --> 00:37:17.037
governance is an important first step.

00:37:17.120 --> 00:37:20.557
Governance and regulation of AI are high

00:37:20.640 --> 00:37:22.957
on Denmark's agenda, including

00:37:23.040 --> 00:37:26.797
regulation of large uh tech companies.

00:37:26.880 --> 00:37:28.637
These efforts are not breaks on

00:37:28.720 --> 00:37:32.077
innovation and growth. They are I think

00:37:32.160 --> 00:37:34.957
essential efforts to ensure that techn

00:37:35.040 --> 00:37:37.517
technological progress serves the public

00:37:37.600 --> 00:37:41.037
interest and humanity in general. The

00:37:41.120 --> 00:37:43.757
unprecedented speed of AI innovation

00:37:43.840 --> 00:37:46.957
demonstrates the need for safeguards to

00:37:47.040 --> 00:37:50.877
ensure AI systems remain safe, remain

00:37:50.960 --> 00:37:53.437
secure and trustworthy.

00:37:53.520 --> 00:37:55.517
uncontrolled new technology could lead

00:37:55.600 --> 00:37:58.717
to dangerous and harmful effects.

00:37:58.800 --> 00:38:01.837
We are already seeing how they pose new

00:38:01.920 --> 00:38:05.037
threats including to those who protects

00:38:05.120 --> 00:38:07.277
civilians in conflict above all the

00:38:07.360 --> 00:38:10.557
medical mission. So Denmark underlines

00:38:10.640 --> 00:38:15.197
uh the importance of ensuring
the responsible use of AI and that the use

00:38:15.280 --> 00:38:16.957
of AI is in accordance with

00:38:17.040 --> 00:38:20.237
international law. In this regard, we

00:38:20.320 --> 00:38:22.237
call on all states to ensure human

00:38:22.320 --> 00:38:25.357
control of AI and accountability for the

00:38:25.440 --> 00:38:28.877
development and use of AI, not least in

00:38:28.960 --> 00:38:31.997
the military domain where accountability

00:38:32.080 --> 00:38:36.797
is absolutely essential. Mr. President,
Denmark will continue to support the

00:38:36.880 --> 00:38:40.957
meaningful engagement of the Security
Council and promote an inclusive

00:38:41.040 --> 00:38:43.197
multistakeholder approach to AI

00:38:43.280 --> 00:38:45.517
governance. Civil society and the

00:38:45.600 --> 00:38:48.637
private sector must play leading roles.

00:38:48.720 --> 00:38:51.837
Science and evidence must always inform

00:38:51.920 --> 00:38:55.357
our policies. We see the independent

00:38:55.440 --> 00:38:57.517
international scientific panel as an

00:38:57.600 --> 00:38:59.757
important contribution to credible and

00:38:59.840 --> 00:39:02.557
evidence-based AI governance. The panel

00:39:02.640 --> 00:39:05.597
provides us with a common scientific

00:39:05.680 --> 00:39:08.717
reference point. So, Mr. President in in

00:39:08.800 --> 00:39:12.157
in in closing, we need to consider how

00:39:12.240 --> 00:39:14.957
AI can contribute positively to shape

00:39:15.040 --> 00:39:16.717
international peace and security and

00:39:16.800 --> 00:39:19.917
bridge global digital divides. At the

00:39:20.000 --> 00:39:23.197
same time, we must remain cleareyed of

00:39:23.280 --> 00:39:25.917
the significant risks and start building

00:39:26.000 --> 00:39:29.357
some strong u guard rails so we can

00:39:29.440 --> 00:39:31.597
manage the global risk in due time.

00:39:31.680 --> 00:39:34.797
Thank you.

00:39:34.880 --> 00:39:37.917
I thank his excellency

00:39:38.000 --> 00:39:40.237
Mr. Esmson for his statement and I now

00:39:40.320 --> 00:39:43.597
give the floor for Carlos

00:39:43.680 --> 00:39:47.757
Minister of Foreign Affairs for Panama.

00:39:47.840 --> 00:39:51.277
>> Thank you, Mr. Minister, Mr. President.

00:39:51.360 --> 00:39:54.077
We thank France for convening this

00:39:54.160 --> 00:39:56.477
important meeting.

00:39:56.560 --> 00:40:00.317
Moreover, we thank Mr. Joshua Benjio,

00:40:00.400 --> 00:40:03.677
co-chair of the Independent Scientific
Panel of the United Nations on

00:40:03.760 --> 00:40:07.277
Artificial Intelligence, Mr. Sam Alman,

00:40:07.360 --> 00:40:11.837
CEO of OpenAI, Mr. Dario Amade, CEO of

00:40:11.920 --> 00:40:16.237
Anthropic, and Mr. Clemon Deong, CEO of

00:40:16.320 --> 00:40:21.080
HuggingFace for their
valuable presentations.

00:40:21.440 --> 00:40:25.037
Although today's meeting invites
us to consider the risks caused and

00:40:25.120 --> 00:40:28.077
exacerbated by artificial intelligence,

00:40:28.160 --> 00:40:30.717
for Panama, it is necessary to begin

00:40:30.800 --> 00:40:33.277
this statement by highlighting its

00:40:33.360 --> 00:40:35.597
potential to transform the world in

00:40:35.680 --> 00:40:37.917
which we live.

00:40:38.000 --> 00:40:40.637
The capacity of these technologies to

00:40:40.720 --> 00:40:43.757
analyze enormous amounts of information,

00:40:43.840 --> 00:40:47.437
identify patterns, generate increasingly

00:40:47.520 --> 00:40:50.157
sophisticated videos, images, and code,

00:40:50.240 --> 00:40:52.717
to automate tasks, and to support

00:40:52.800 --> 00:40:55.517
planning and decision-making processes

00:40:55.600 --> 00:41:00.077
are all evidence of a transformation
that is already underway, the scope of

00:41:00.160 --> 00:41:02.317
which we are barely beginning to

00:41:02.400 --> 00:41:06.397
understand. These tools have a wide

00:41:06.480 --> 00:41:08.397
array of applications that can benefit

00:41:08.480 --> 00:41:11.517
humanity. In Panama, our own national

00:41:11.600 --> 00:41:16.317
artificial intelligence strategy
is identifying opportunities in sectors

00:41:16.400 --> 00:41:17.997
such as healthcare, agriculture,

00:41:18.080 --> 00:41:20.717
education, and logistics where these

00:41:20.800 --> 00:41:23.597
technologies can offer solutions which

00:41:23.680 --> 00:41:25.997
just a few years ago would have seemed

00:41:26.080 --> 00:41:28.077
unimaginable.

00:41:28.160 --> 00:41:32.477
And it is specifically in this area of

00:41:32.560 --> 00:41:35.357
progress where if applied to other

00:41:35.440 --> 00:41:37.517
purposes, we also see the main risks

00:41:37.600 --> 00:41:39.677
emerging. Artificial intelligence can

00:41:39.760 --> 00:41:42.397
also be used to amplify disinformation

00:41:42.480 --> 00:41:45.517
campaigns to allow for cyber attacks

00:41:45.600 --> 00:41:48.157
against critical infrastructure and to

00:41:48.240 --> 00:41:53.037
facilitate access to knowledge for the
development of weapons among other uses

00:41:53.120 --> 00:41:56.957
that are already a threat
to international security. And these risks

00:41:57.040 --> 00:41:58.877
are not hypothetical scenarios. We are

00:41:58.960 --> 00:42:03.517
beginning to see them come to pass in

00:42:03.600 --> 00:42:07.357
situations that are directly related
to this council. There are recent reports

00:42:07.440 --> 00:42:12.477
on the use of artificial intelligence by
non-state actors to develop guidance and

00:42:12.560 --> 00:42:14.237
navigation systems for rockets and

00:42:14.320 --> 00:42:16.957
missiles. On top of that, there are

00:42:17.040 --> 00:42:19.997
reports of autonomous weapons being used

00:42:20.080 --> 00:42:21.757
that are able to navigate and select

00:42:21.840 --> 00:42:24.717
targets completely independently or with

00:42:24.800 --> 00:42:27.917
very limited human intervention.

00:42:28.000 --> 00:42:30.637
It is precisely this growing autonomy

00:42:30.720 --> 00:42:35.597
which raises one of the most serious
concerns that was noted in the concept

00:42:35.680 --> 00:42:39.117
note for this meeting. That is the
possibility that the evolution of these

00:42:39.200 --> 00:42:43.837
systems could exceed our ability
to maintain effective human control of

00:42:43.920 --> 00:42:47.677
them. There are decisions
which by their very

00:42:47.760 --> 00:42:50.317
nature cannot be reduced to a

00:42:50.400 --> 00:42:52.237
statistical result, a statistical

00:42:52.320 --> 00:42:54.877
outcome. When human lives are at stake,

00:42:54.960 --> 00:42:59.677
it is vital that ethical, legal, and
contextual considerations be taken into

00:42:59.760 --> 00:43:01.757
account which no calculation or

00:43:01.840 --> 00:43:05.837
algorithmic system can replace.

00:43:05.920 --> 00:43:08.477
This concern for control is not limited

00:43:08.560 --> 00:43:12.237
to extreme scenarios. As we begin to use

00:43:12.320 --> 00:43:15.917
analysis produced by artificial
intelligence to steer and guide our

00:43:16.000 --> 00:43:20.077
decisions, it is essential that those
that are in leadership positions uh be

00:43:20.160 --> 00:43:24.957
able to understand how these tools have
arrived at a certain recommendation,

00:43:25.040 --> 00:43:26.637
what information underpins that

00:43:26.720 --> 00:43:28.557
recommendation, and what the margin of

00:43:28.640 --> 00:43:32.557
error is. Mr. President Recognizing the

00:43:32.640 --> 00:43:34.557
risks of artificial intelligence does

00:43:34.640 --> 00:43:37.837
not mean ignoring the benefits that this

00:43:37.920 --> 00:43:39.997
technology is already providing our

00:43:40.080 --> 00:43:43.277
societies. The challenge is finding a

00:43:43.360 --> 00:43:47.997
balance that allows us to take advantage
of these opportunities while also

00:43:48.080 --> 00:43:52.237
responding to the risks with uh
responsibly. In that regard, Panama

00:43:52.320 --> 00:43:56.157
recognizes the important role
of the International Scientific Panel on

00:43:56.240 --> 00:43:58.477
artificial intelligence to provide

00:43:58.560 --> 00:44:03.277
impartial and scientific information
that can guide the discussions of the

00:44:03.360 --> 00:44:06.637
international community
on these technologies.

00:44:06.720 --> 00:44:09.757
Equally, we welcome the initiative of

00:44:09.840 --> 00:44:12.637
the tech sector actors here present to

00:44:12.720 --> 00:44:14.877
proactively share information about

00:44:14.960 --> 00:44:17.277
specific cases of the misuse of these

00:44:17.360 --> 00:44:20.477
technologies and also about incidents in

00:44:20.560 --> 00:44:22.797
which their growing autonomy has

00:44:22.880 --> 00:44:25.837
generated concerns about the possible

00:44:25.920 --> 00:44:29.357
loss of human control of them.

00:44:29.440 --> 00:44:31.357
President,

00:44:31.440 --> 00:44:33.197
>> this meeting is evidence that in order

00:44:33.280 --> 00:44:36.717
to address the range of risks related to

00:44:36.800 --> 00:44:41.757
artificial intelligence, closer
cooperation between states, the tech

00:44:41.840 --> 00:44:43.677
sector, and other relevant actors will

00:44:43.760 --> 00:44:46.397
be essential. To that end, it is

00:44:46.480 --> 00:44:48.797
essential that we establish mechanisms

00:44:48.880 --> 00:44:51.757
for a more structured and permanent

00:44:51.840 --> 00:44:53.997
exchange of information on emerging

00:44:54.080 --> 00:44:57.437
risks, incidents, misuse, and good

00:44:57.520 --> 00:45:02.077
practices. artificial
intelligence will inevitably

00:45:02.160 --> 00:45:03.997
continue to evolve and continue to

00:45:04.080 --> 00:45:07.117
transform the world in which we live

00:45:07.200 --> 00:45:11.837
and the testimonies that we've
heard today are evidence of that. Our

00:45:11.920 --> 00:45:13.997
responsibility is to make sure that this

00:45:14.080 --> 00:45:16.637
progress continues to be in the service

00:45:16.720 --> 00:45:19.117
of people and that it contributes to the

00:45:19.200 --> 00:45:21.357
construction of safer, more peaceful,

00:45:21.440 --> 00:45:26.157
more prosperous societies with a
perspective of equity that allows all of

00:45:26.240 --> 00:45:30.317
the nations of the world including the
global south to participate and benefit

00:45:30.400 --> 00:45:33.117
from its development.
Thank you very much. Merc

00:45:33.200 --> 00:45:35.997
>> I thank his excellency
for his statement

00:45:36.080 --> 00:45:38.957
and I now give the floor to

00:45:39.040 --> 00:45:43.677
his excellency Christo menopolis deputy
minister for digital governance and AI

00:45:43.760 --> 00:45:46.717
of Greece. >> Mr.
President, I wish to begin by

00:45:46.800 --> 00:45:51.677
thanking France along with the briefers
for their remarks and for bringing to

00:45:51.760 --> 00:45:55.837
the attention of the security council
the crucial issue of the link between

00:45:55.920 --> 00:45:57.757
the rapid technological advances in the

00:45:57.840 --> 00:46:00.477
field of AI and international peace and

00:46:00.560 --> 00:46:03.117
security. A subject that we will also

00:46:03.200 --> 00:46:06.717
tackle during our presidency. AI is

00:46:06.800 --> 00:46:08.637
already changing the way we think about

00:46:08.720 --> 00:46:11.357
international peace and security. It can

00:46:11.440 --> 00:46:14.237
help us prevent conflict, improve early

00:46:14.320 --> 00:46:17.037
warning, and make UN operations more

00:46:17.120 --> 00:46:19.677
effective. But it also create risks that

00:46:19.760 --> 00:46:22.397
we are only beginning to understand. For

00:46:22.480 --> 00:46:25.277
Greece, our starting point is clear. AI

00:46:25.360 --> 00:46:28.157
must remain humanentric, secure, and

00:46:28.240 --> 00:46:31.037
trustworthy. It development and use must

00:46:31.120 --> 00:46:33.037
respect international law including the

00:46:33.120 --> 00:46:38.157
UN charter, international humanitarian
law and international human rights law

00:46:38.240 --> 00:46:41.837
and human responsibility and
accountability must remain at the center

00:46:41.920 --> 00:46:44.557
of our approach. Our national approach

00:46:44.640 --> 00:46:47.277
combines innovation with governance.

00:46:47.360 --> 00:46:49.277
Greece has established the special

00:46:49.360 --> 00:46:52.317
secretariat for AI and data governance,

00:46:52.400 --> 00:46:54.957
implemented the national law for the EU

00:46:55.040 --> 00:46:57.517
AI act and is investing in strategic

00:46:57.600 --> 00:46:59.357
infrastructure including the

00:46:59.440 --> 00:47:03.037
supercomputer and the Faros AI factory.

00:47:03.120 --> 00:47:06.317
In April 2025 together with France and

00:47:06.400 --> 00:47:08.557
the Republic of Korea, Greece organized

00:47:08.640 --> 00:47:10.877
an area formula meeting on harnessing

00:47:10.960 --> 00:47:13.917
safe, inclusive and trustworthy AI for

00:47:14.000 --> 00:47:18.317
the maintenance of international peace
and security. Today's discussion give us

00:47:18.400 --> 00:47:20.477
an opportunity to take that conversation

00:47:20.560 --> 00:47:24.157
further. Allow me to make three points.

00:47:24.240 --> 00:47:26.957
First, we need to better understand the

00:47:27.040 --> 00:47:30.477
risks. AI can process enormous amounts

00:47:30.560 --> 00:47:32.957
of information and help us make better

00:47:33.040 --> 00:47:35.677
informed decisions. But speed is not

00:47:35.760 --> 00:47:38.317
always an advantage. In situations of

00:47:38.400 --> 00:47:40.557
crisis, AI can compress the time

00:47:40.640 --> 00:47:42.797
available for decisions and encourage

00:47:42.880 --> 00:47:46.077
excessive reliance on automated outputs.

00:47:46.160 --> 00:47:48.797
As AI increasingly enters decision-m

00:47:48.880 --> 00:47:50.957
processes, we must ensure that greater

00:47:51.040 --> 00:47:53.597
speed does not come at the expense of

00:47:53.680 --> 00:47:56.637
sound human judgment. These concerns

00:47:56.720 --> 00:47:59.357
become even more serious when AI enters

00:47:59.440 --> 00:48:02.237
military decision-m. Technology can

00:48:02.320 --> 00:48:04.877
assist human judgment. It cannot replace

00:48:04.960 --> 00:48:07.437
it. Responsibility and accountability

00:48:07.520 --> 00:48:09.677
for critical decisions must always

00:48:09.760 --> 00:48:12.957
remain human. Nowhere in is this more

00:48:13.040 --> 00:48:16.157
important than in the nuclear domain. We

00:48:16.240 --> 00:48:18.237
must be extremely cautious about not

00:48:18.320 --> 00:48:20.877
integrating AI into nuclear command and

00:48:20.960 --> 00:48:26.157
control. The risk of miscalculation
and unintended escalation are simply too

00:48:26.240 --> 00:48:29.757
great. Decisions of such consequence must

00:48:29.840 --> 00:48:32.397
remain under human control.

00:48:32.480 --> 00:48:35.037
Second, we need practical ways to reduce

00:48:35.120 --> 00:48:38.157
these risks. This starts with dialogue.

00:48:38.240 --> 00:48:42.237
We need greater transparency,
confidence building, and a better common

00:48:42.320 --> 00:48:46.477
understanding of how international law
applies to the development and use of

00:48:46.560 --> 00:48:50.877
AI. Important work
is already under way. The

00:48:50.960 --> 00:48:52.717
reaim process and the political

00:48:52.800 --> 00:48:55.037
declaration responsible military use of

00:48:55.120 --> 00:48:58.317
AI and autonomy have helped advance the

00:48:58.400 --> 00:49:03.357
international discussion on responsible
military uses of AI. The United Nations

00:49:03.440 --> 00:49:07.357
has an essential role in bringing
these different efforts together.

00:49:07.440 --> 00:49:09.677
The Security Council also has a role to

00:49:09.760 --> 00:49:12.637
play not in regulating technology and

00:49:12.720 --> 00:49:14.957
not in duplicating the work of other UN

00:49:15.040 --> 00:49:17.917
bodies. Our responsibility is to

00:49:18.000 --> 00:49:20.157
understand what these technologies mean

00:49:20.240 --> 00:49:22.477
for international peace and security and

00:49:22.560 --> 00:49:25.277
to act when those implications fall

00:49:25.360 --> 00:49:28.957
within our mandate. Third, addressing

00:49:29.040 --> 00:49:30.797
risk should not mean holding back

00:49:30.880 --> 00:49:33.357
innovation. We should make use of the

00:49:33.440 --> 00:49:36.237
opportunities AI provides while putting

00:49:36.320 --> 00:49:38.957
appropriate safeguards in place and we

00:49:39.040 --> 00:49:40.797
cannot have this discussion without

00:49:40.880 --> 00:49:43.277
those who develop and use these

00:49:43.360 --> 00:49:46.157
technologies. We therefore support close

00:49:46.240 --> 00:49:49.117
engagement with academia, civil society

00:49:49.200 --> 00:49:51.837
and the private sector alongside states

00:49:51.920 --> 00:49:54.557
and the United Nations. Such engagement

00:49:54.640 --> 00:49:57.037
should be inclusive, geographically

00:49:57.120 --> 00:50:00.397
representative and gender balanced. Mr.

00:50:00.480 --> 00:50:03.037
President, AI will increasingly affect

00:50:03.120 --> 00:50:05.757
international peace and security. We

00:50:05.840 --> 00:50:08.317
cannot stop technological change, nor

00:50:08.400 --> 00:50:13.037
should we seek to, but we can work
together to ensure that human beings

00:50:13.120 --> 00:50:14.957
remain responsible for decisions that

00:50:15.040 --> 00:50:18.237
matter most. For Greece, this means

00:50:18.320 --> 00:50:22.317
keeping international law, human
responsibility, and accountability at

00:50:22.400 --> 00:50:24.557
the center of our approach. It also

00:50:24.640 --> 00:50:27.517
means keeping this council focused on

00:50:27.600 --> 00:50:31.837
its primarily responsibility,
the maintenance of international peace and

00:50:31.920 --> 00:50:34.317
security. So Greece will continue to

00:50:34.400 --> 00:50:36.157
work with all members of the council

00:50:36.240 --> 00:50:40.317
towards this objective. Thank you.

00:50:40.400 --> 00:50:43.037
>> I thank his excellency Mr. Demenopulos

00:50:43.120 --> 00:50:45.357
for his statement. I now give the floor

00:50:45.440 --> 00:50:48.557
to his excellency Mr. Ali Muhammad Omar,

00:50:48.640 --> 00:50:53.560
state minister for foreign
affairs of Somalia.

00:50:55.920 --> 00:50:57.997
Mr. President, I thank France for

00:50:58.080 --> 00:51:01.037
convening this timely debate. I also

00:51:01.120 --> 00:51:04.877
thank Mr. Banio, Oldman, Almadeo, and

00:51:04.960 --> 00:51:08.557
Dingua for their insightful briefing.

00:51:08.640 --> 00:51:10.477
Artificial intelligence is no longer

00:51:10.560 --> 00:51:12.957
only a technological or a developmental

00:51:13.040 --> 00:51:16.557
issue. is our current reality. Rapidly

00:51:16.640 --> 00:51:19.197
transforming society,

00:51:19.280 --> 00:51:21.837
economies and our approach to peace and

00:51:21.920 --> 00:51:24.717
security. It is role in a cyber

00:51:24.800 --> 00:51:27.277
operationist disinformation, critical

00:51:27.360 --> 00:51:29.677
infrastructure and military decision

00:51:29.760 --> 00:51:32.077
making places it firmly within the

00:51:32.160 --> 00:51:34.397
United Nations mandate.

00:51:34.480 --> 00:51:36.797
We must prevent AI from becoming a

00:51:36.880 --> 00:51:39.997
driver of a conflict while actively

00:51:40.080 --> 00:51:43.277
supporting it responsible use in early

00:51:43.360 --> 00:51:46.077
warning, crisis prevention, peacekeeping

00:51:46.160 --> 00:51:51.117
and humanitarian action and civilian
protection. In that vein, four core

00:51:51.200 --> 00:51:52.637
principles

00:51:52.720 --> 00:51:56.557
should guide us. Number one, first a

00:51:56.640 --> 00:51:58.637
human responsibility must remain

00:51:58.720 --> 00:52:02.077
absolute. AI must operate strictly

00:52:02.160 --> 00:52:03.997
within the framework of international

00:52:04.080 --> 00:52:06.957
law, the purposes and the principles of

00:52:07.040 --> 00:52:09.197
the UN charter and the respect of her

00:52:09.280 --> 00:52:11.757
human rights and her human dignity.

00:52:11.840 --> 00:52:14.957
States cannot delegate legal, ethical

00:52:15.040 --> 00:52:18.877
and moral accountability
to an algorithmist.

00:52:18.960 --> 00:52:21.277
We welcome the declaration affirming

00:52:21.360 --> 00:52:23.517
that AI must remain under human

00:52:23.600 --> 00:52:28.237
direction, insight and control.

00:52:28.320 --> 00:52:30.557
This principle is indispensable in

00:52:30.640 --> 00:52:33.837
decision involving the use of force,

00:52:33.920 --> 00:52:36.877
autonomous weapon system and nuclear

00:52:36.960 --> 00:52:40.797
command and control. Second, safety must

00:52:40.880 --> 00:52:44.077
must come before a deployment.

00:52:44.160 --> 00:52:46.957
Recent incidents involving AI agent

00:52:47.040 --> 00:52:50.157
reportedly bypassing safeguards.

00:52:50.240 --> 00:52:52.877
Concealing their actionist and obtaining

00:52:52.960 --> 00:52:54.557
unauthorized

00:52:54.640 --> 00:52:58.077
access raises a serious question about

00:52:58.160 --> 00:53:02.957
whether existing controllers can keep
base with the increasing autonomous

00:53:03.040 --> 00:53:06.797
systems. Private companies
develop and control

00:53:06.880 --> 00:53:09.517
many of these systems. But commercial

00:53:09.600 --> 00:53:12.237
incentives cannot determine how much

00:53:12.320 --> 00:53:16.717
risky a human must accept. We therefore

00:53:16.800 --> 00:53:20.397
support a mandatory incident reporting,

00:53:20.480 --> 00:53:23.997
layered technical safeguards, independent

00:53:24.080 --> 00:53:26.397
audits and measures to mitigate

00:53:26.480 --> 00:53:29.517
algorithmic bias and protect vulnerable

00:53:29.600 --> 00:53:31.197
populations.

00:53:31.280 --> 00:53:33.677
When agreed capability threshold are

00:53:33.760 --> 00:53:36.957
crossed, deployment must be suspended

00:53:37.040 --> 00:53:41.117
until risks can be reliably managed. The

00:53:41.200 --> 00:53:43.677
Secretary General's proposal to explore

00:53:43.760 --> 00:53:46.717
an international institution capable of

00:53:46.800 --> 00:53:50.317
setting standards enabling verification

00:53:50.400 --> 00:53:52.797
and convening states when such threshold

00:53:52.880 --> 00:53:55.677
are crossed deserves a careful

00:53:55.760 --> 00:53:59.277
consideration. Third, governance must

00:53:59.360 --> 00:54:02.717
address Africa's specific vulnerability

00:54:02.800 --> 00:54:05.917
and ensure equitable access. Limited

00:54:06.000 --> 00:54:08.397
computing infrastructure, unreliable

00:54:08.480 --> 00:54:11.357
connectivity, shortages of specialized

00:54:11.440 --> 00:54:14.717
expert and fragmented regulatory

00:54:14.800 --> 00:54:18.157
capacity constrain the the ability of

00:54:18.240 --> 00:54:21.517
many African states to evaluate advanced

00:54:21.600 --> 00:54:24.717
systems. models train it on a foreign

00:54:24.800 --> 00:54:28.237
data may misrepresent African language

00:54:28.320 --> 00:54:31.997
and context while synthetic synthetic

00:54:32.080 --> 00:54:36.317
data can inflame a social tensionist

00:54:36.400 --> 00:54:39.197
interfere in a fragile be processes and

00:54:39.280 --> 00:54:41.517
undermine elections.

00:54:41.600 --> 00:54:44.717
Africa cannot remain merely a market for

00:54:44.800 --> 00:54:48.157
system systems developed elsewhere, a

00:54:48.240 --> 00:54:50.797
source of a road data or a testing

00:54:50.880 --> 00:54:54.397
ground for a technology it did not help

00:54:54.480 --> 00:54:57.917
govern. We must ver we must prevent new

00:54:58.000 --> 00:55:01.197
forms of technological dependency and

00:55:01.280 --> 00:55:05.757
digital digital colonaliz colonialism

00:55:05.840 --> 00:55:08.157
by supporting a regional framework

00:55:08.240 --> 00:55:10.877
particularly partnership with African

00:55:10.960 --> 00:55:13.597
Union and strategies like the ban

00:55:13.680 --> 00:55:16.077
African AI initiative.

00:55:16.160 --> 00:55:18.557
Developing countries require a targeted

00:55:18.640 --> 00:55:21.357
international cooperation, financing to

00:55:21.440 --> 00:55:24.157
bridge the digital and infrastructure,

00:55:24.240 --> 00:55:27.437
energy and skills divide alongside

00:55:27.520 --> 00:55:30.077
access to local data sets, evaluation

00:55:30.160 --> 00:55:33.517
tools and computing power. Ensuring

00:55:33.600 --> 00:55:35.997
every nation has the right to develop,

00:55:36.080 --> 00:55:41.277
choose and assess AI technology
is essential for a national digital

00:55:41.360 --> 00:55:45.037
sovereignty sovereignty
and our shared global

00:55:45.120 --> 00:55:48.157
stability. Fourth, the United Nation

00:55:48.240 --> 00:55:51.517
must serve as the main global

00:55:51.600 --> 00:55:54.157
as the main general for global govern AI

00:55:54.240 --> 00:55:58.557
governance. Somali supporters
international normies anchored in the

00:55:58.640 --> 00:56:01.117
universal legitimacy of the UN rather

00:56:01.200 --> 00:56:03.677
than fragmented rules established by

00:56:03.760 --> 00:56:06.557
small number of technological advanced

00:56:06.640 --> 00:56:09.117
states and private cooperations that

00:56:09.200 --> 00:56:11.677
risk reinforcing existing global

00:56:11.760 --> 00:56:13.757
inequalities.

00:56:13.840 --> 00:56:17.037
Global governance must uphold a people

00:56:17.120 --> 00:56:20.397
centered approach that directs AI toward

00:56:20.480 --> 00:56:23.357
human well-being and common good.

00:56:23.440 --> 00:56:26.557
developing regional regionist Africa in

00:56:26.640 --> 00:56:29.117
particular must must be granted a

00:56:29.200 --> 00:56:32.237
genuine representation and a voice in

00:56:32.320 --> 00:56:35.277
these multilateral processes

00:56:35.360 --> 00:56:37.757
alongside the institution institutional

00:56:37.840 --> 00:56:40.157
and technological capacity to evaluate

00:56:40.240 --> 00:56:43.597
AI systems independently

00:56:43.680 --> 00:56:45.997
rather than relying on external

00:56:46.080 --> 00:56:48.317
determination. The measures of our

00:56:48.400 --> 00:56:51.437
success will not be the sophistication

00:56:51.520 --> 00:56:54.157
of our declaration but whether we

00:56:54.240 --> 00:56:57.757
prevent the first AI enabled crisis from

00:56:57.840 --> 00:56:59.517
becoming the next international

00:56:59.600 --> 00:57:03.277
conflict. The power of AI intelligence

00:57:03.360 --> 00:57:06.877
does not entail inevitable outcome. It

00:57:06.960 --> 00:57:09.757
shaped and and and given meaning by our

00:57:09.840 --> 00:57:12.397
values and our choices. The technology

00:57:12.480 --> 00:57:15.037
is advancing quickly. Our collective

00:57:15.120 --> 00:57:17.997
responsibility must advance faster. I

00:57:18.080 --> 00:57:20.637
thank you.

00:57:20.720 --> 00:57:23.117
>> I thank his excellency Mr. Omar for his

00:57:23.200 --> 00:57:26.157
statement. I now give the floor to his

00:57:26.240 --> 00:57:30.877
excellency Mr. Ed Milliband, Minister
of Foreign Affairs uh minister, secretary

00:57:30.960 --> 00:57:32.797
of state foreign, commonwealth
and development affairs of the United

00:57:32.880 --> 00:57:37.677
Kingdom. uh can I really thank you
for convening this very very important uh

00:57:37.760 --> 00:57:41.997
session and I also want to put on record
my thanks to the expert speakers uh who

00:57:42.080 --> 00:57:44.477
have provided such an illuminating uh

00:57:44.560 --> 00:57:47.677
contribution uh to this session.

00:57:47.760 --> 00:57:49.437
I think we should stand back and take

00:57:49.520 --> 00:57:53.117
stock of what confronts us uh in this

00:57:53.200 --> 00:57:55.997
session of the security council because

00:57:56.080 --> 00:57:58.557
AI is a technology which undoubtedly has

00:57:58.640 --> 00:58:00.557
the power to transform our world for the

00:58:00.640 --> 00:58:03.037
better. I'm sure we all see it in our

00:58:03.120 --> 00:58:05.997
daily lives. We see it in all of our

00:58:06.080 --> 00:58:08.717
countries. But today, the founders of

00:58:08.800 --> 00:58:11.677
this technology are coming to us with a

00:58:11.760 --> 00:58:15.437
stark warning. They are telling us not

00:58:15.520 --> 00:58:18.317
just that they relish the chance of this

00:58:18.400 --> 00:58:22.397
technology to do good, but they fear its

00:58:22.480 --> 00:58:25.757
ability to do harm.

00:58:25.840 --> 00:58:29.197
They are telling us in stark terms that

00:58:29.280 --> 00:58:32.477
it cannot just be left to them to

00:58:32.560 --> 00:58:35.837
prevent disaster from befalling the

00:58:35.920 --> 00:58:37.437
world.

00:58:37.520 --> 00:58:41.357
I say today we cannot ignore them

00:58:41.440 --> 00:58:44.317
because they are right

00:58:44.400 --> 00:58:46.317
because we cannot abdicate our

00:58:46.400 --> 00:58:49.437
responsibility to our citizens. Because

00:58:49.520 --> 00:58:52.077
we cannot outsource to private companies

00:58:52.160 --> 00:58:55.517
the first duty of government to protect

00:58:55.600 --> 00:58:59.917
our people whatever the issue.

00:59:00.000 --> 00:59:03.837
This security council was founded
to address the great collective challenges

00:59:03.920 --> 00:59:06.877
of our time. So it is for governments,

00:59:06.960 --> 00:59:11.517
for governments, as many of the people
who have made contributions today from

00:59:11.600 --> 00:59:15.517
the tech world have said, for all
of us around tables such as this to work

00:59:15.600 --> 00:59:19.917
together and together with industry
to deliver in the interests of the people

00:59:20.000 --> 00:59:22.237
that have sent us here and that is what

00:59:22.320 --> 00:59:25.277
the UK government is committed to doing.

00:59:25.360 --> 00:59:27.757
We must start with safety. This means

00:59:27.840 --> 00:59:29.997
working harder to ensure frontier AI

00:59:30.080 --> 00:59:33.117
models are rigorously tested and

00:59:33.200 --> 00:59:35.437
building a system of assurance around

00:59:35.520 --> 00:59:37.517
model development to allow us to

00:59:37.600 --> 00:59:41.997
establish trust that models are safe.

00:59:42.080 --> 00:59:44.317
Second, transparency.

00:59:44.400 --> 00:59:49.037
The companies building the models cannot
simply be left to their own devices.

00:59:49.120 --> 00:59:53.677
Governments must have the visibility
to understand and assess what AI companies

00:59:53.760 --> 00:59:55.117
are doing.

00:59:55.200 --> 00:59:57.757
The leading AI companies have actually

00:59:57.840 --> 01:00:00.797
committed to provide this visibility.

01:00:00.880 --> 01:00:05.597
It's really really important and it
is an offer that we and they should follow

01:00:05.680 --> 01:00:07.437
through on.

01:00:07.520 --> 01:00:10.557
Third, resilience. We have to harden our

01:00:10.640 --> 01:00:13.357
defenses to reduce the impact of AI

01:00:13.440 --> 01:00:15.597
enabled incidents.

01:00:15.680 --> 01:00:18.397
Even if we test every model, we must be

01:00:18.480 --> 01:00:22.397
prepared for AI to be misused. That

01:00:22.480 --> 01:00:24.317
includes boosting our cyber defenses,

01:00:24.400 --> 01:00:26.397
protecting critical infrastructure, and

01:00:26.480 --> 01:00:30.397
reinforcing our financial systems.

01:00:30.480 --> 01:00:34.157
And we must support our populations to
manage and adjust to the disruption that

01:00:34.240 --> 01:00:38.797
AI will bring to how we live and work,
as well as to harness its incredible

01:00:38.880 --> 01:00:42.797
benefits. I say today
that the UK will play its

01:00:42.880 --> 01:00:47.517
full part in this process.
We established the first AI security

01:00:47.600 --> 01:00:50.717
institute which shares its leading
expertise, testing capabilities and

01:00:50.800 --> 01:00:55.357
research working handin glove with the
United States and the leading labs on AI

01:00:55.440 --> 01:00:57.517
safety. As our prime minister set out

01:00:57.600 --> 01:01:00.317
yesterday, the UK will put AI at the

01:01:00.400 --> 01:01:03.917
heart of its G20 presidency next year.

01:01:04.000 --> 01:01:08.557
Building on our track record of bringing
the right people together on this issue,

01:01:08.640 --> 01:01:13.037
breaking down silos between government,
industry, researchers, and civil

01:01:13.120 --> 01:01:15.517
society. delivering on the potential of

01:01:15.600 --> 01:01:19.357
AI for good and dealing with the risks

01:01:19.440 --> 01:01:22.877
whether in terms of the future of jobs,
protecting children, cyber security,

01:01:22.960 --> 01:01:25.917
biocurity, national security or

01:01:26.000 --> 01:01:28.877
financial stability

01:01:28.960 --> 01:01:31.677
as we also work to agreeing to working

01:01:31.760 --> 01:01:36.397
towards agreeing a set of a single set
of global principles and standards to

01:01:36.480 --> 01:01:38.157
ensure the transparency and access

01:01:38.240 --> 01:01:41.437
necessary to increase our preparedness

01:01:41.520 --> 01:01:44.717
to ensure that AI development is safe

01:01:44.800 --> 01:01:47.997
and as the previous speaker said to

01:01:48.080 --> 01:01:50.637
seize potential for the global south and

01:01:50.720 --> 01:01:54.317
global north alike. Let me just end on

01:01:54.400 --> 01:01:58.077
this uh point. I think overall this is

01:01:58.160 --> 01:02:00.717
not a choice between progress and

01:02:00.800 --> 01:02:05.357
safety. We are only going to be able
to innovate and capture the extraordinary

01:02:05.440 --> 01:02:09.197
benefits of AI which I'm sure we
all share in this room if people have

01:02:09.280 --> 01:02:13.837
confidence that the right safeguards
are in place. So together we must take

01:02:13.920 --> 01:02:15.997
whatever meas measures are necessary so

01:02:16.080 --> 01:02:20.157
that AI is secure, safe and trustworthy

01:02:20.240 --> 01:02:22.157
while still advancing its incredible

01:02:22.240 --> 01:02:25.197
potential to drive growth, cure disease,

01:02:25.280 --> 01:02:28.637
and transform people's life chances. I

01:02:28.720 --> 01:02:31.597
end on this point finally.

01:02:31.680 --> 01:02:34.557
We can't say we haven't been warned.

01:02:34.640 --> 01:02:37.117
We have been warned by the people who

01:02:37.200 --> 01:02:39.437
have founded this incredible technology

01:02:39.520 --> 01:02:43.197
with its huge potential for good.

01:02:43.280 --> 01:02:46.477
We must act on their warnings and we

01:02:46.560 --> 01:02:49.117
must act on their warnings together.

01:02:49.200 --> 01:02:51.277
Thank you.

01:02:51.360 --> 01:02:53.357
>> I thank his excellency Mr.
Milliband for

01:02:53.440 --> 01:02:56.237
his statement. I now give the floor to

01:02:56.320 --> 01:03:01.037
his excellency Mr. Michael Katzio,
assistant to the president and director

01:03:01.120 --> 01:03:03.117
of the White House Office of Science and

01:03:03.200 --> 01:03:07.277
Technology Policy, United States.

01:03:07.360 --> 01:03:09.437
>> My sincerest thanks to the Republic of

01:03:09.520 --> 01:03:12.557
France as president of the Council.

01:03:12.640 --> 01:03:17.357
Distinguished colleagues, I am grateful
for another opportunity to discuss the

01:03:17.440 --> 01:03:19.197
most consequential technological

01:03:19.280 --> 01:03:21.677
development of our time, super

01:03:21.760 --> 01:03:25.837
intelligence, formerly known as AI. Our

01:03:25.920 --> 01:03:28.317
primary responsibility is to ensure that

01:03:28.400 --> 01:03:31.037
continued advances in super intelligence

01:03:31.120 --> 01:03:33.757
benefit our citizens and safeguard their

01:03:33.840 --> 01:03:37.437
liberty and way of life. The frontier of

01:03:37.520 --> 01:03:41.037
intelligence is advancing rapidly. That

01:03:41.120 --> 01:03:43.437
is not a reason to pause its further

01:03:43.520 --> 01:03:46.717
development or to constrain it with new

01:03:46.800 --> 01:03:50.317
global governance structures.

01:03:50.400 --> 01:03:54.317
As the leader in frontier super
intelligence, the the US has taken

01:03:54.400 --> 01:03:58.717
actions to ensure we are able
to keep the frontier moving safely and

01:03:58.800 --> 01:04:02.717
responsibly. We've engaged
Frontier Labs on testing

01:04:02.800 --> 01:04:06.077
and evaluation of new model capabilities,

01:04:06.160 --> 01:04:07.917
established cyber security clearing

01:04:08.000 --> 01:04:10.237
houses for coordination on vulnerability

01:04:10.320 --> 01:04:14.957
mitigation, and clarified the existing
law enforcement tools available to

01:04:15.040 --> 01:04:19.197
address harmful uses
of super intelligence.

01:04:19.280 --> 01:04:23.277
Internationally, the United States
is committed to diplomacy and dialogue on

01:04:23.360 --> 01:04:26.077
emerging technologies, but international

01:04:26.160 --> 01:04:29.117
dialogue in this forum and in others

01:04:29.200 --> 01:04:34.077
cannot be allowed to drift towards
global governance. As President Trump

01:04:34.160 --> 01:04:38.157
said before the General Assembly
yesterday, the United States totally

01:04:38.240 --> 01:04:43.117
rejects any attempt to construct
a globalist scheme of control of super

01:04:43.200 --> 01:04:47.677
intelligence. Only weeks ago,
G20 innovation ministers

01:04:47.760 --> 01:04:50.077
reached consensus on the Carolina

01:04:50.160 --> 01:04:52.477
principles, agreeing that countries

01:04:52.560 --> 01:04:54.877
should preserve national sovereignty in

01:04:54.960 --> 01:04:57.677
the governance of emerging technologies,

01:04:57.760 --> 01:05:02.077
not abdicate responsibility
to international bodies.

01:05:02.160 --> 01:05:06.077
The American people's representatives
will legislate and regulate on the

01:05:06.160 --> 01:05:08.637
American people's behalf. You should do

01:05:08.720 --> 01:05:11.677
the same for your people.

01:05:11.760 --> 01:05:15.837
Only through an open cooperative
ecosystem built on sovereign

01:05:15.920 --> 01:05:20.797
infrastructure and domestic capacity
can the world realize this technology's

01:05:20.880 --> 01:05:24.477
potential. US international
efforts seek to build

01:05:24.560 --> 01:05:27.677
that capacity by sharing our technology

01:05:27.760 --> 01:05:31.037
and providing the tools for adoption.

01:05:31.120 --> 01:05:35.197
We want you, our allies and partners
to share in the benefits from this

01:05:35.280 --> 01:05:37.517
technological revolution.

01:05:37.600 --> 01:05:39.677
That is the purpose of our exports

01:05:39.760 --> 01:05:42.557
program. The US is the only technology

01:05:42.640 --> 01:05:44.957
superpower willing and able to truly

01:05:45.040 --> 01:05:49.917
empower partner nations in your pursuit
of a sovereign super intelligence

01:05:50.000 --> 01:05:54.077
capability. American
companies will work with you to

01:05:54.160 --> 01:05:57.917
build gold standard secure
infrastructure and help you implement

01:05:58.000 --> 01:06:00.797
the applications and development the

01:06:00.880 --> 01:06:05.117
tools and capacity you
need to serve your people.

01:06:05.200 --> 01:06:07.277
In this era, technology sovereignty is

01:06:07.360 --> 01:06:10.557
realized in technology adoption. If you

01:06:10.640 --> 01:06:14.957
take the risks and benefits of super
intelligence seriously, the most

01:06:15.040 --> 01:06:17.437
important thing you can do is to develop

01:06:17.520 --> 01:06:20.157
your country's ability to understand and

01:06:20.240 --> 01:06:23.677
use it. Deploy it widely. Build

01:06:23.760 --> 01:06:26.157
expertise inside your government.

01:06:26.240 --> 01:06:27.837
Prepare your institutions and

01:06:27.920 --> 01:06:30.157
regulators. Understand what these

01:06:30.240 --> 01:06:33.677
systems can and cannot do. You cannot

01:06:33.760 --> 01:06:38.237
govern a technology you do not understand.
This body and others like it

01:06:38.320 --> 01:06:40.637
should focus on sharing best practices

01:06:40.720 --> 01:06:43.437
to build domestic capacity, not

01:06:43.520 --> 01:06:46.877
establishing a global regulatory scheme.

01:06:46.960 --> 01:06:51.597
It is through a policy of permissionless
innovation, a tireless pursuit of the

01:06:51.680 --> 01:06:55.837
frontier, and an optimism about
the future of technology that the United

01:06:55.920 --> 01:06:57.837
States built the greatest innovation

01:06:57.920 --> 01:07:01.997
ecosystem ever known to man. For 250

01:07:02.080 --> 01:07:04.077
years, the American character has sought

01:07:04.160 --> 01:07:06.717
a virtuous balance between innovation

01:07:06.800 --> 01:07:10.397
and safety, independence and security.

01:07:10.480 --> 01:07:14.317
Super intelligence is the latest
in that long series of challenges and

01:07:14.400 --> 01:07:16.557
opportunities. In the service of

01:07:16.640 --> 01:07:18.717
humankind, it can become the greatest

01:07:18.800 --> 01:07:21.357
engine of growth and prosperity in

01:07:21.440 --> 01:07:24.397
history. A prosperous future will not be

01:07:24.480 --> 01:07:27.037
secured by a global regulator. It will

01:07:27.120 --> 01:07:29.197
be secured by sovereign nations that

01:07:29.280 --> 01:07:31.837
adopt super intelligence, responsible

01:07:31.920 --> 01:07:34.477
companies that build it and free people

01:07:34.560 --> 01:07:40.760
who refuse to be ruled by fear. Thank you.

01:07:41.360 --> 01:07:43.597
>> I thank his excellency Mr. Katius for

01:07:43.680 --> 01:07:46.797
his statement. I now give the floor to

01:07:46.880 --> 01:07:49.357
his excellency Mr. Sheikh Abdullahbin

01:07:49.440 --> 01:07:51.677
Ali al- Khalifa, acting under secretary

01:07:51.760 --> 01:07:57.360
for political affairs in the ministry
foreign affairs of Bahrain.

01:07:58.160 --> 01:08:01.277
Mr. President, allow me

01:08:01.360 --> 01:08:05.597
at the outset to express sincere thanks
and appreciation to the friendly French

01:08:05.680 --> 01:08:07.677
Republic for convening this important

01:08:07.760 --> 01:08:10.477
session. I also extend my thanks to Mr.

01:08:10.560 --> 01:08:14.237
Benjio, co-chair of the UN Independent

01:08:14.320 --> 01:08:16.077
Scientific Panel on Artificial

01:08:16.160 --> 01:08:18.557
Intelligence,

01:08:18.640 --> 01:08:22.957
Mr. Altman, CEO of Open AI, Mr. Mr.

01:08:23.040 --> 01:08:28.317
Amode, CEO of Anthropic and Mr. Dong,

01:08:28.400 --> 01:08:31.437
CEO of Hugging Face, for their valuable

01:08:31.520 --> 01:08:36.077
briefings. Mr. President, discussing the

01:08:36.160 --> 01:08:38.397
implications of artificial intelligence

01:08:38.480 --> 01:08:40.717
is no longer a matter of intellectual

01:08:40.800 --> 01:08:43.117
interest. It has become an urgent

01:08:43.200 --> 01:08:45.917
necessity dictated by this council's

01:08:46.000 --> 01:08:48.077
responsibilities for maintaining

01:08:48.160 --> 01:08:50.797
international peace and security. Guided

01:08:50.880 --> 01:08:53.117
by this principle, the kingdom of

01:08:53.200 --> 01:08:55.757
Bahrain emphasized in the statement

01:08:55.840 --> 01:08:57.837
delivered by his royal highness, Prince

01:08:57.920 --> 01:09:00.877
Salman bin Hammad al- Khalifa, Crown

01:09:00.960 --> 01:09:03.357
Prince and Prime Minister during the

01:09:03.440 --> 01:09:06.397
79th session of the General Assembly.

01:09:06.480 --> 01:09:11.037
The importance of concluding
an international treaty to govern the

01:09:11.120 --> 01:09:13.117
development of artificial intelligence

01:09:13.200 --> 01:09:15.517
and ensure that it serves as a force for

01:09:15.600 --> 01:09:18.317
innovation rather than destruction.

01:09:18.400 --> 01:09:20.957
Consistent with this approach, Bahrain

01:09:21.040 --> 01:09:23.597
recently had the honor of co-hosting an

01:09:23.680 --> 01:09:25.357
interactive dialogue with France and

01:09:25.440 --> 01:09:28.157
Latvia on harnessing this technology for

01:09:28.240 --> 01:09:31.677
conflict resolution. Mr. President, we

01:09:31.760 --> 01:09:33.997
stress that the rapid evolution of

01:09:34.080 --> 01:09:36.477
artificial intelligence expands the

01:09:36.560 --> 01:09:39.197
scope of security threats, particularly

01:09:39.280 --> 01:09:41.757
when these technologies are exploited by

01:09:41.840 --> 01:09:45.837
terrorists and criminal groups.

01:09:45.920 --> 01:09:48.637
As the autonomy of these systems grows,

01:09:48.720 --> 01:09:51.517
we emphasize the need to maintain human

01:09:51.600 --> 01:09:53.997
oversight and accountability at the core

01:09:54.080 --> 01:09:56.397
of decisions with serious security

01:09:56.480 --> 01:09:58.877
implications and ensure their full

01:09:58.960 --> 01:10:01.197
compliance with international law and

01:10:01.280 --> 01:10:04.397
international humanitarian law.

01:10:04.480 --> 01:10:07.757
Unlike automated algorithms, only humans

01:10:07.840 --> 01:10:10.397
exercise reason, responsibility, and

01:10:10.480 --> 01:10:12.797
conscience and possess the ability to

01:10:12.880 --> 01:10:15.917
discern complex contexts. At the same

01:10:16.000 --> 01:10:19.197
time, we must not overlook the immense

01:10:19.280 --> 01:10:21.517
opportunities to harness artificial

01:10:21.600 --> 01:10:23.917
intelligence to enhance early warning

01:10:24.000 --> 01:10:26.957
systems, monitor indicators of tension,

01:10:27.040 --> 01:10:29.517
support mediation efforts, and improve

01:10:29.600 --> 01:10:32.237
effective peacekeeping operations. We

01:10:32.320 --> 01:10:34.637
also stress the need for bridging the

01:10:34.720 --> 01:10:39.517
digital divide and facilitating
technology transfer to ensure that all

01:10:39.600 --> 01:10:41.117
nations can benefit from these

01:10:41.200 --> 01:10:44.317
capabilities. In conclusion, we stress

01:10:44.400 --> 01:10:46.637
the need for the Security Council to

01:10:46.720 --> 01:10:49.597
hold to hold more regular and systematic

01:10:49.680 --> 01:10:52.397
discussions on artificial intelligence.

01:10:52.480 --> 01:10:56.957
These discussions should complement
broader UN efforts to address frameworks

01:10:57.040 --> 01:10:58.877
for the development and testing of these

01:10:58.960 --> 01:11:02.077
models in line with international law.

01:11:02.160 --> 01:11:04.637
We stand ready to contribute to

01:11:04.720 --> 01:11:08.077
advancing these discussions to build a

01:11:08.160 --> 01:11:12.637
responsible and balanced
international approach. One

01:11:12.720 --> 01:11:14.637
that places humans at the heart of

01:11:14.720 --> 01:11:17.437
decision making and ensures that the

01:11:17.520 --> 01:11:20.717
innovations we create today continue to

01:11:20.800 --> 01:11:23.917
serve the fe the peace we all strive to

01:11:24.000 --> 01:11:27.597
uphold in this council. Thank you, Mr.

01:11:27.680 --> 01:11:30.680
President.

01:11:31.760 --> 01:11:34.877
I thank his excellency al Khalifa for

01:11:34.960 --> 01:11:40.600
his statement and I now give the floor
to the representative of Latvia

01:11:40.800 --> 01:11:43.597
president. >> Thank you Mr. President.

01:11:43.680 --> 01:11:47.277
>> First of all I would like to thank
the French presidency for convening this

01:11:47.360 --> 01:11:50.397
meeting on an important and extremely

01:11:50.480 --> 01:11:52.237
urgent [music] subject for international

01:11:52.320 --> 01:11:54.397
security. briefers professor Yosha

01:11:54.480 --> 01:11:57.837
Benjer Sam Alman Dario Amade and Clemon

01:11:57.920 --> 01:12:01.117
Deong for the enlightening assessment

01:12:01.200 --> 01:12:03.117
and also the proposals for safe

01:12:03.200 --> 01:12:06.797
development of the AI.

01:12:06.880 --> 01:12:10.077
uh president, the rapid pace of AI

01:12:10.160 --> 01:12:12.397
development is already having profound

01:12:12.480 --> 01:12:15.517
impact across all domains, including

01:12:15.600 --> 01:12:18.477
international peace and security.

01:12:18.560 --> 01:12:19.837
Alongside the extraordinary

01:12:19.920 --> 01:12:23.197
opportunities that AI offers, we also

01:12:23.280 --> 01:12:25.677
see growing risks manifesting both in

01:12:25.760 --> 01:12:29.357
peace time and in conflict settings.

01:12:29.440 --> 01:12:32.797
This is a moment for collective action

01:12:32.880 --> 01:12:35.357
as we hear today and as highlighted in

01:12:35.440 --> 01:12:37.757
the call for control of frontier AI

01:12:37.840 --> 01:12:41.037
models initiated by Norway and Finland

01:12:41.120 --> 01:12:44.557
issued on Monday 2 by sec 22nd world

01:12:44.640 --> 01:12:46.397
leaders including the president of

01:12:46.480 --> 01:12:49.357
Latvia.

01:12:49.440 --> 01:12:52.077
We believe that the security council has

01:12:52.160 --> 01:12:54.237
a responsibility to ensure that the

01:12:54.320 --> 01:12:56.637
implications of AI for international

01:12:56.720 --> 01:12:59.517
peace and security are fully considered

01:12:59.600 --> 01:13:01.997
without duplicating work undertaken in

01:13:02.080 --> 01:13:04.477
other four of course but today's

01:13:04.560 --> 01:13:06.637
discussion demonstrates the council's

01:13:06.720 --> 01:13:09.997
ability to convene member state science

01:13:10.080 --> 01:13:12.477
and industry leaders to address this

01:13:12.560 --> 01:13:16.237
pressing matter. President,

01:13:16.320 --> 01:13:20.797
two distinct sources of AI related
risks for peace and security are becoming

01:13:20.880 --> 01:13:22.317
evident.

01:13:22.400 --> 01:13:25.517
Both demand our attention. First are the

01:13:25.600 --> 01:13:28.317
risks stemming from the misuse of AI by

01:13:28.400 --> 01:13:31.197
state and non-state actors

01:13:31.280 --> 01:13:34.637
in includes application of AI in malign

01:13:34.720 --> 01:13:38.557
cyber activities, information
manipulation and even attempts to

01:13:38.640 --> 01:13:41.117
acquire destructive capabilities such as

01:13:41.200 --> 01:13:43.277
biological weapons.

01:13:43.360 --> 01:13:46.317
Second, a more novel and complex

01:13:46.400 --> 01:13:48.877
challenge as outlined by the briefers

01:13:48.960 --> 01:13:51.677
and the recent reports stems from the

01:13:51.760 --> 01:13:54.797
technology itself such as misalignment

01:13:54.880 --> 01:13:57.597
and risk of loss of control. Such

01:13:57.680 --> 01:13:59.917
behavior risks are source of new forms

01:14:00.000 --> 01:14:02.077
of vulnerability

01:14:02.160 --> 01:14:04.797
including to critical infrastructure.

01:14:04.880 --> 01:14:07.357
Alarmingly, recent incidents indicate

01:14:07.440 --> 01:14:11.837
that not not only can AI systems
act contrary to human intentions and

01:14:11.920 --> 01:14:16.877
interests, but they are becoming better
at manipulating and deceiving humans as

01:14:16.960 --> 01:14:20.797
they do so.
These challenges are not an isolated

01:14:20.880 --> 01:14:23.837
matter as they add to the already

01:14:23.920 --> 01:14:25.597
complex international peace and security

01:14:25.680 --> 01:14:29.037
landscape. actions by misaligned AI may

01:14:29.120 --> 01:14:32.077
fuel further tensions and miscalculation

01:14:32.160 --> 01:14:34.637
due to the difficulty to track, verify,

01:14:34.720 --> 01:14:37.677
and attribute them. The increasing use

01:14:37.760 --> 01:14:42.397
of AI and autonomous systems
in the military domain poses additional

01:14:42.480 --> 01:14:46.797
dilemmas, particularly where
reliability, control, and escalation

01:14:46.880 --> 01:14:49.117
risks intersect.

01:14:49.200 --> 01:14:51.837
President, responsible action starts at

01:14:51.920 --> 01:14:54.477
the industry level, but also with

01:14:54.560 --> 01:14:59.117
governance, clear guard rails,
and capable institutions.

01:14:59.200 --> 01:15:01.357
A multistakeholder approach is essential

01:15:01.440 --> 01:15:05.037
to manage risks and build trust in AI

01:15:05.120 --> 01:15:10.157
through transparency and accountability
through establishment of adequate safety

01:15:10.240 --> 01:15:14.637
protocols for the development and use
of AI systems through independent

01:15:14.720 --> 01:15:16.397
evaluation

01:15:16.480 --> 01:15:18.077
through connecting the government's

01:15:18.160 --> 01:15:21.917
mandate academic expertise civil society

01:15:22.000 --> 01:15:25.997
and industry innovation and all
these require international

01:15:26.080 --> 01:15:29.997
cooperation coordination norms.
setting exchange of

01:15:30.080 --> 01:15:34.877
best practices and capacity building
is key to bridge the divide and foster

01:15:34.960 --> 01:15:38.797
interoperability.
The convening of the UN global dialogue

01:15:38.880 --> 01:15:41.117
on AI governance and the establishment

01:15:41.200 --> 01:15:43.117
of the independent scientific panel on

01:15:43.200 --> 01:15:46.077
AI are important steps.

01:15:46.160 --> 01:15:50.557
Our further efforts should keep up with
the pace of development of frontier AI

01:15:50.640 --> 01:15:54.077
technologies as highlighted
in the call for control

01:15:54.160 --> 01:15:56.717
of frontier AI models. We need an

01:15:56.800 --> 01:15:58.957
open-minded approach by the UN member

01:15:59.040 --> 01:16:01.757
states to build on current international

01:16:01.840 --> 01:16:04.317
mechanisms and explore developing new

01:16:04.400 --> 01:16:07.357
ones. The overall objective should be to

01:16:07.440 --> 01:16:11.597
ensure that technology serves people,
safeguards fundamental rights and

01:16:11.680 --> 01:16:14.237
freedoms and delivers benefits that do

01:16:14.320 --> 01:16:18.717
not come at the expense
of uncontrollable risks. If there is one

01:16:18.800 --> 01:16:20.317
thing that we should be able to agree

01:16:20.400 --> 01:16:22.957
on, AI must remain under human

01:16:23.040 --> 01:16:26.317
direction, oversight, and control.

01:16:26.400 --> 01:16:29.997
President, uh, considering its role
in maintaining international peace and

01:16:30.080 --> 01:16:33.997
security, this council must remain
actively engaged in assessing the

01:16:34.080 --> 01:16:36.157
implications of AI for global peace and

01:16:36.240 --> 01:16:41.037
security. With this technology become
becoming increasingly relevant to the

01:16:41.120 --> 01:16:43.117
council's work in the area of conflict

01:16:43.200 --> 01:16:45.997
prevention and resolution, proactive

01:16:46.080 --> 01:16:48.557
approach is necessary. and building on

01:16:48.640 --> 01:16:51.277
the recent council's inform informal in

01:16:51.360 --> 01:16:54.237
interactive dialogue on on AI that or

01:16:54.320 --> 01:16:59.037
was organized by Latia and France and
supported by Bahrain and Greece. We are

01:16:59.120 --> 01:17:03.117
committed to taking further concrete
steps. For instance, establishment of

01:17:03.200 --> 01:17:05.277
council's informal expert group that

01:17:05.360 --> 01:17:07.677
would enable more in-depth council

01:17:07.760 --> 01:17:12.797
analysis on the imp of the impact of AI
related technologies on international

01:17:12.880 --> 01:17:16.477
peace and security and the
implementation of the council's

01:17:16.560 --> 01:17:19.437
mandates. Latia will continue to work

01:17:19.520 --> 01:17:21.997
with interested partners on the council

01:17:22.080 --> 01:17:26.557
in the coming month. I thank you.

01:17:26.640 --> 01:17:30.317
I thank the permanent representative of
Latvia for her statement. I now give the

01:17:30.400 --> 01:17:36.120
floor to the permanent
representative of Colombia.

01:17:36.800 --> 01:17:40.520
>> Thank you, Mr. President.

01:17:41.280 --> 01:17:44.280
Colombia

01:17:45.920 --> 01:17:48.077
considers

01:17:48.160 --> 01:17:50.717
that here.

01:17:50.800 --> 01:17:53.757
We are not discussing the the benefits

01:17:53.840 --> 01:17:56.717
of artificial intelligence but um really

01:17:56.800 --> 01:17:59.117
it is the risks that come to the four

01:17:59.200 --> 01:18:03.037
the uncontrollable in of unavoidable

01:18:03.120 --> 01:18:06.557
risks and quite clearly following the

01:18:06.640 --> 01:18:13.160
briefings that we have
heard the risk for humanity

01:18:13.600 --> 01:18:17.197
>> we cannot ignore
bluntly the report that

01:18:17.280 --> 01:18:21.997
we have heard is the first time this
Security Council and is on the record

01:18:22.080 --> 01:18:23.917
for history

01:18:24.000 --> 01:18:27.117
has heard a case where machines commit

01:18:27.200 --> 01:18:29.117
crimes.

01:18:29.200 --> 01:18:33.437
This is not just a random mistake.

01:18:33.520 --> 01:18:35.997
Other person if will be other company

01:18:36.080 --> 01:18:39.357
even state secrets please try to convey

01:18:39.440 --> 01:18:44.717
the idea for those that produce
intelligence or artificial intelligence

01:18:44.800 --> 01:18:46.317
that

01:18:46.400 --> 01:18:49.437
the large part of the planet does not.

01:18:49.520 --> 01:18:52.157
And as such the critical infrastructure

01:18:52.240 --> 01:18:55.117
including cyber security systems will be

01:18:55.200 --> 01:18:56.877
exposed.

01:18:56.960 --> 01:18:59.917
If you read the cases that were

01:19:00.000 --> 01:19:03.917
explained to us today, we will know that

01:19:04.000 --> 01:19:06.157
it will be a crime even against the

01:19:06.240 --> 01:19:09.277
state if this will take place at a state

01:19:09.360 --> 01:19:13.437
level and it can happen.

01:19:13.520 --> 01:19:16.557
The second element uh for us and perhaps

01:19:16.640 --> 01:19:19.757
the major concern is that we haven't

01:19:19.840 --> 01:19:23.277
framed the issue correctly

01:19:23.360 --> 01:19:28.717
and we must make a refrain of the issue.

01:19:28.800 --> 01:19:31.277
I do not think and Colombia does not

01:19:31.360 --> 01:19:34.397
think that the issue here is to limit

01:19:34.480 --> 01:19:36.717
human innovation

01:19:36.800 --> 01:19:39.997
but machine innovation.

01:19:40.080 --> 01:19:44.840
And I explain my words here. Different

01:19:45.680 --> 01:19:48.637
reports has telling us that we face two

01:19:48.720 --> 01:19:52.077
risks. One is misalignment and the other

01:19:52.160 --> 01:19:55.037
is the loss of control. when it comes to

01:19:55.120 --> 01:19:58.637
misalignment is that the machines are

01:19:58.720 --> 01:20:03.437
separating or are finding ways to cheat

01:20:03.520 --> 01:20:07.277
break the instructions that were given

01:20:07.360 --> 01:20:11.917
by humans that is the agents AI agents

01:20:12.000 --> 01:20:14.317
are able to do this repeatedly

01:20:14.400 --> 01:20:16.157
systematically

01:20:16.240 --> 01:20:20.397
and uh so far even in the cyber security

01:20:20.480 --> 01:20:23.837
case that we analyze they were able to

01:20:23.920 --> 01:20:27.437
reach internet and finds way ways to

01:20:27.520 --> 01:20:29.597
cheat the system. Other times they are

01:20:29.680 --> 01:20:32.317
able to find out passwords, send

01:20:32.400 --> 01:20:35.597
thousands of messages even against the

01:20:35.680 --> 01:20:38.157
company that was testing these uh

01:20:38.240 --> 01:20:41.597
machines in the sense box.

01:20:41.680 --> 01:20:43.917
But the other one and the one that

01:20:44.000 --> 01:20:45.757
should be very very concerning for this

01:20:45.840 --> 01:20:50.317
council is the loss of control.

01:20:50.400 --> 01:20:53.517
So we would like to go with you with a

01:20:53.600 --> 01:20:55.917
logical narrative.

01:20:56.000 --> 01:21:01.277
We are telling us today that AI has an

01:21:01.360 --> 01:21:03.597
immense capability to storage

01:21:03.680 --> 01:21:07.357
information that is far far away from

01:21:07.440 --> 01:21:09.757
human capability.

01:21:09.840 --> 01:21:12.797
We are knowing now that given the models

01:21:12.880 --> 01:21:16.397
that we are adapting but also the design

01:21:16.480 --> 01:21:20.077
of the algorithm and the paths that we

01:21:20.160 --> 01:21:22.477
open in that design.

01:21:22.560 --> 01:21:27.117
These capabilities also allow to involve

01:21:27.200 --> 01:21:30.557
a different type of reasoning each day

01:21:30.640 --> 01:21:34.237
more complex interdisciplinary.

01:21:34.320 --> 01:21:38.237
In other words, soon there won't be any

01:21:38.320 --> 01:21:41.837
human capable of following the

01:21:41.920 --> 01:21:44.717
operation, the development of what we

01:21:44.800 --> 01:21:46.797
are creating.

01:21:46.880 --> 01:21:49.837
In the first days of the models, we

01:21:49.920 --> 01:21:52.397
notice how the machine were able to pass

01:21:52.480 --> 01:21:55.277
a bar exam in three months, a medical

01:21:55.360 --> 01:21:57.997
license exam in three months. So there

01:21:58.080 --> 01:22:00.317
are no humans that we know that can do

01:22:00.400 --> 01:22:03.117
that. But at this state-of-the-art of

01:22:03.200 --> 01:22:06.317
the technology, we are facing a

01:22:06.400 --> 01:22:09.117
different type of situation

01:22:09.200 --> 01:22:14.317
which is a pass from human mistakes to

01:22:14.400 --> 01:22:18.477
machines random mistakes and we have to

01:22:18.560 --> 01:22:21.597
be aware of the risks that we are facing

01:22:21.680 --> 01:22:23.757
and assuming if that is the path we

01:22:23.840 --> 01:22:25.997
choose to take.

01:22:26.080 --> 01:22:29.597
You see the critical issue here is to

01:22:29.680 --> 01:22:32.717
know as a security council

01:22:32.800 --> 01:22:35.677
what are we regulating,

01:22:35.760 --> 01:22:39.117
why we need to regulate it and why we

01:22:39.200 --> 01:22:41.997
should do it.

01:22:42.080 --> 01:22:44.957
Most importantly, Colombia understands

01:22:45.040 --> 01:22:47.517
the position of the United States and

01:22:47.600 --> 01:22:50.237
shares partially

01:22:50.320 --> 01:22:54.637
its practical approach which is

01:22:54.720 --> 01:22:59.837
an international body oversign
the sovereign production of AI is

01:22:59.920 --> 01:23:03.997
impractical for two basic
reasons. There are of

01:23:04.080 --> 01:23:06.397
course constitutional and national

01:23:06.480 --> 01:23:10.157
security reasons in every country. So

01:23:10.240 --> 01:23:13.517
the legal battle will be overwhelming.

01:23:13.600 --> 01:23:16.557
But secondly, if one country stops, the

01:23:16.640 --> 01:23:19.197
other says I do not. Especially those

01:23:19.280 --> 01:23:22.397
that are leading the race. For the rest

01:23:22.480 --> 01:23:26.237
of humanity, the concerns
and for those that are

01:23:26.320 --> 01:23:29.037
behind but will soon catch up. The

01:23:29.120 --> 01:23:33.917
concern is what we do next. And that's

01:23:34.000 --> 01:23:37.757
where our country can perhaps offer a

01:23:37.840 --> 01:23:40.317
different approach.

01:23:40.400 --> 01:23:42.877
And again, no between an inelectable

01:23:42.960 --> 01:23:46.717
choice between safety and development,

01:23:46.800 --> 01:23:48.957
but perhaps

01:23:49.040 --> 01:23:52.157
a way to make sure

01:23:52.240 --> 01:23:56.317
that all humans on Earth are included.

01:23:56.400 --> 01:23:59.757
Not just by shedding or as we use the

01:23:59.840 --> 01:24:03.757
term adapting our countries because it

01:24:03.840 --> 01:24:06.797
seems to us in the global south that

01:24:06.880 --> 01:24:11.197
when we call about adapting is to buy a

01:24:11.280 --> 01:24:14.477
product that we will not produce

01:24:14.560 --> 01:24:17.357
and once we buy the product we can adapt

01:24:17.440 --> 01:24:19.757
our structures and infrastructures.

01:24:19.840 --> 01:24:22.877
Again, partially true, but many of these

01:24:22.960 --> 01:24:26.557
countries will face an undeniable gap

01:24:26.640 --> 01:24:28.797
that will extend even a century between

01:24:28.880 --> 01:24:31.837
another country and the other if we do

01:24:31.920 --> 01:24:36.077
not participate of the production of AI.

01:24:36.160 --> 01:24:38.877
That's one issue and that's a very very

01:24:38.960 --> 01:24:42.077
important issue for the global south. We

01:24:42.160 --> 01:24:45.677
do not believe that just by adapting our

01:24:45.760 --> 01:24:49.917
infrastructure, by making it safer, we

01:24:50.000 --> 01:24:53.517
will also get rid of of the lateral and

01:24:53.600 --> 01:24:56.717
collateral problems. And that's when we

01:24:56.800 --> 01:25:00.397
come to the point today we could lose

01:25:00.480 --> 01:25:03.997
millions of jobs. That is a possibility

01:25:04.080 --> 01:25:06.557
that is being entertained even in

01:25:06.640 --> 01:25:09.837
developed countries. We have seen how AI

01:25:09.920 --> 01:25:12.877
start writing the code

01:25:12.960 --> 01:25:16.557
and that for us is perhaps the only

01:25:16.640 --> 01:25:19.037
place that we should start the

01:25:19.120 --> 01:25:20.637
regulation

01:25:20.720 --> 01:25:24.797
because no one discusses no one would

01:25:24.880 --> 01:25:27.357
say that AI is not beneficial for

01:25:27.440 --> 01:25:30.237
research, health development,

01:25:30.320 --> 01:25:32.877
technology, communications,

01:25:32.960 --> 01:25:36.557
uh cure diseases etc. Well, the benefits

01:25:36.640 --> 01:25:39.437
are wide now and we are not trying to

01:25:39.520 --> 01:25:42.077
stop it. But it seems that we are

01:25:42.160 --> 01:25:45.757
reaching a new step on AI and the US is

01:25:45.840 --> 01:25:48.797
right. This is super intelligence for

01:25:48.880 --> 01:25:52.317
one particular reason only. We are about

01:25:52.400 --> 01:25:56.797
to take the step that a
greater and potentially

01:25:56.880 --> 01:25:59.117
more complex technology

01:25:59.200 --> 01:26:01.837
uh quantum technology will boost

01:26:01.920 --> 01:26:04.957
artificial intelligence in the design in

01:26:05.040 --> 01:26:07.597
the model in the modeling and the

01:26:07.680 --> 01:26:12.157
development. So the risks
that we are facing today I

01:26:12.240 --> 01:26:14.077
want to call the attention of the the

01:26:14.160 --> 01:26:17.277
council are pre-quantum risks. So that

01:26:17.360 --> 01:26:19.997
is before we reach that that state but

01:26:20.080 --> 01:26:21.917
we are very very close to reach that

01:26:22.000 --> 01:26:25.757
state. But the concerning side for us

01:26:25.840 --> 01:26:29.037
and perhaps this is what we will offer

01:26:29.120 --> 01:26:31.117
as a different approach rather than

01:26:31.200 --> 01:26:35.517
thinking of an international oversight.

01:26:35.600 --> 01:26:37.437
The Security Council is concerned

01:26:37.520 --> 01:26:40.317
because this technology has the

01:26:40.400 --> 01:26:43.997
potential to disrupt and create not just

01:26:44.080 --> 01:26:46.317
more inequality across the planet

01:26:46.400 --> 01:26:50.397
including violence but also

01:26:50.480 --> 01:26:53.517
development gaps that will lead us to

01:26:53.600 --> 01:26:57.357
other types of conflicts including

01:26:57.440 --> 01:26:59.437
global migration

01:26:59.520 --> 01:27:03.757
etc. So we propose to you, we submit to

01:27:03.840 --> 01:27:06.477
you uh an important consideration in

01:27:06.560 --> 01:27:08.797
which perhaps we can start just for that

01:27:08.880 --> 01:27:12.317
domain because this is it. This is the

01:27:12.400 --> 01:27:14.797
moment.

01:27:14.880 --> 01:27:17.917
Autonomous design and autonomous

01:27:18.000 --> 01:27:20.317
development. This is what is really

01:27:20.400 --> 01:27:23.357
really at stake here.

01:27:23.440 --> 01:27:27.357
Should we allow that machines,

01:27:27.440 --> 01:27:30.157
AI agents take the next step that are

01:27:30.240 --> 01:27:32.157
showing by the experiments that are

01:27:32.240 --> 01:27:34.797
shown by the cases they present to us

01:27:34.880 --> 01:27:37.197
and that we that had been published

01:27:37.280 --> 01:27:39.837
actually not just today. There are many

01:27:39.920 --> 01:27:42.877
article journals these are everywhere

01:27:42.960 --> 01:27:45.437
that tell us how the the machine finds

01:27:45.520 --> 01:27:47.597
another way a different way different

01:27:47.680 --> 01:27:50.557
than humans. One will lose control. One

01:27:50.640 --> 01:27:52.717
of the major risk of risk management

01:27:52.800 --> 01:27:56.397
here which is once we do no longer

01:27:56.480 --> 01:27:59.997
understand what the machine is doing

01:28:00.080 --> 01:28:03.757
because if we go to super intelligence

01:28:03.840 --> 01:28:06.717
and we do not possess as humans such a

01:28:06.800 --> 01:28:09.437
capacity, we are assuming that we are

01:28:09.520 --> 01:28:13.437
controlling something that we cannot uh

01:28:13.520 --> 01:28:17.757
understand. And one of the main elements

01:28:17.840 --> 01:28:20.797
for this technology and I listen to the

01:28:20.880 --> 01:28:23.357
all interventions today.

01:28:23.440 --> 01:28:26.157
Uh we want to make sure that this remain

01:28:26.240 --> 01:28:29.837
human. Well, you see that's the problem.

01:28:29.920 --> 01:28:33.197
The human brain needs to understand

01:28:33.280 --> 01:28:35.677
the information that has in front of it

01:28:35.760 --> 01:28:38.877
to make a decision. But it seems to us

01:28:38.960 --> 01:28:43.277
that we are soon going to leave that

01:28:43.360 --> 01:28:45.517
stage and we are going to leave that

01:28:45.600 --> 01:28:48.237
stage because the way the technology is

01:28:48.320 --> 01:28:51.197
being developed and this is not just a

01:28:51.280 --> 01:28:53.517
matter of ethical limits. We understand

01:28:53.600 --> 01:28:57.277
that that will be a contradictory for

01:28:57.360 --> 01:29:00.397
many and it will be also a problem in

01:29:00.480 --> 01:29:03.437
terms of legislation and regulations.

01:29:03.520 --> 01:29:06.157
Some might even see that as a red tape

01:29:06.240 --> 01:29:11.757
etc. No, this is another step because

01:29:11.840 --> 01:29:13.997
if the machine is the one that starts

01:29:14.080 --> 01:29:17.277
leading the design, in other words, the

01:29:17.360 --> 01:29:20.957
methods, the paths, the outcomes,

01:29:21.040 --> 01:29:24.237
if this is not just and soon will not be

01:29:24.320 --> 01:29:26.477
a problem of

01:29:26.560 --> 01:29:31.197
the input and the unreachable goal. In

01:29:31.280 --> 01:29:33.117
other words, it seems now that the

01:29:33.200 --> 01:29:35.917
machine is trying to just get the goal

01:29:36.000 --> 01:29:39.597
done, get things done. And to do that is

01:29:39.680 --> 01:29:43.277
finding paths that humans usually do not

01:29:43.360 --> 01:29:46.797
take one and did not instruct to take

01:29:46.880 --> 01:29:50.557
yet it find a way and it cheats and it

01:29:50.640 --> 01:29:52.797
conceals the outcome

01:29:52.880 --> 01:29:55.757
and that is a concern of course. One is

01:29:55.840 --> 01:29:58.317
that is misalignment. The second one now

01:29:58.400 --> 01:30:01.437
however is that

01:30:01.520 --> 01:30:05.437
soon and Mr. Sam Oldman really really

01:30:05.520 --> 01:30:08.797
pointed out to this risk is that the

01:30:08.880 --> 01:30:11.197
autonomous development of artificial

01:30:11.280 --> 01:30:14.157
intelligence what really means is this

01:30:14.240 --> 01:30:16.797
that machines write the code that

01:30:16.880 --> 01:30:19.277
machines find the paths that machines

01:30:19.360 --> 01:30:22.157
establish solutions and again this is

01:30:22.240 --> 01:30:26.797
pre-quantum era so it will be naive at

01:30:26.880 --> 01:30:30.717
least to say to think that we will be

01:30:30.800 --> 01:30:32.957
able to even regulate something that

01:30:33.040 --> 01:30:36.077
soon we will not understand.

01:30:36.160 --> 01:30:39.117
And this is why perhaps this is the

01:30:39.200 --> 01:30:42.317
moment in history that all nations has

01:30:42.400 --> 01:30:44.877
to come together at least with one

01:30:44.960 --> 01:30:48.957
common goal. Let's keep it human.

01:30:49.040 --> 01:30:52.957
So we should agree
that we want development.

01:30:53.040 --> 01:30:56.317
We might even agree that
we don't want oversight.

01:30:56.400 --> 01:31:00.717
We should understand the the concerns
in terms of sovereignity especially of the

01:31:00.800 --> 01:31:03.277
nations that have a greater scientific

01:31:03.360 --> 01:31:05.837
development. It is not their go their

01:31:05.920 --> 01:31:09.197
fault to be fully innovated capacities

01:31:09.280 --> 01:31:11.277
etc. But at least for the rest of

01:31:11.360 --> 01:31:14.157
humanity it should be important that we

01:31:14.240 --> 01:31:17.997
decide how far we go. This is again is

01:31:18.080 --> 01:31:22.157
is not an just an ethical country.
We are not talking even about yet about

01:31:22.240 --> 01:31:24.957
human rights. Some things some some

01:31:25.040 --> 01:31:27.357
countries think that we should in terms

01:31:27.440 --> 01:31:29.757
of private etc. No no no no this is the

01:31:29.840 --> 01:31:32.957
next step are we going to allow machines

01:31:33.040 --> 01:31:35.677
to build machines to to design their own

01:31:35.760 --> 01:31:39.037
AI and if we are there already and and

01:31:39.120 --> 01:31:41.757
and that's perhaps the most important

01:31:41.840 --> 01:31:44.317
element is for the countries to decide

01:31:44.400 --> 01:31:49.117
at least that because in practice

01:31:49.200 --> 01:31:52.920
uh Mr. president

01:31:55.920 --> 01:31:58.317
>> for many other countries

01:31:58.400 --> 01:32:01.437
particularly Latin American countries

01:32:01.520 --> 01:32:03.837
in practice

01:32:03.920 --> 01:32:08.797
this means a risk for the national
production of many countries countries

01:32:08.880 --> 01:32:12.957
that that aren't producing artificial
intelligence and those that have been

01:32:13.040 --> 01:32:15.677
invited to buy artificial intelligence

01:32:15.760 --> 01:32:19.837
and invest in it and probably Doing that

01:32:19.920 --> 01:32:24.637
would be the most logical path for them
to modernize the infrastructure of each

01:32:24.720 --> 01:32:26.317
nation and make it as safe and secure as

01:32:26.400 --> 01:32:29.277
possible. But ultimately

01:32:29.360 --> 01:32:32.157
the question remains the same. If we

01:32:32.240 --> 01:32:36.317
have um um regulations even down to the

01:32:36.400 --> 01:32:40.317
production of a tire or a toothbrush, if

01:32:40.400 --> 01:32:44.717
we regulate the production of cars,

01:32:44.800 --> 01:32:48.557
if we um do that with nuclear weapons,

01:32:48.640 --> 01:32:51.597
the question remains the same

01:32:51.680 --> 01:32:54.957
in scientific progress including among

01:32:55.040 --> 01:32:58.957
the most important things the Einstein's

01:32:59.040 --> 01:33:01.517
uh the movement from from uh towards

01:33:01.600 --> 01:33:03.677
Einstein's special theory of relativity.

01:33:03.760 --> 01:33:07.677
We not only verified that time moves
differently in different spaces, but we

01:33:07.760 --> 01:33:10.237
also came to understand the uh

01:33:10.320 --> 01:33:14.717
scientific that scientific precision
is based on perception. And we changed the

01:33:14.800 --> 01:33:18.957
formula from constant velocity that did
not include gravity to another one that

01:33:19.040 --> 01:33:23.037
did include gravity. And we reached
a scientific understanding perhaps the

01:33:23.120 --> 01:33:26.557
most important one of the 20th century
without which we would not have GPS for

01:33:26.640 --> 01:33:30.957
example. But there was one
characteristic common

01:33:31.040 --> 01:33:33.997
to all of those cases, including the
production of nuclear weapons. And that

01:33:34.080 --> 01:33:36.397
is that human beings were always in

01:33:36.480 --> 01:33:41.517
control. Human beings
understood the risks. They

01:33:41.600 --> 01:33:45.197
knew what they were doing and they knew
what they were producing and that's why

01:33:45.280 --> 01:33:47.517
they regulated it.

01:33:47.600 --> 01:33:50.077
The concern

01:33:50.160 --> 01:33:53.517
for countries such as Colombia

01:33:53.600 --> 01:33:57.037
isn't scientific progress. Colombia

01:33:57.120 --> 01:33:59.037
invites investment in artificial

01:33:59.120 --> 01:34:00.797
intelligence

01:34:00.880 --> 01:34:04.477
and wants data centers in its country.

01:34:04.560 --> 01:34:08.477
We are a strategic partner of various

01:34:08.560 --> 01:34:12.877
countries uh including the United
States in particular and naturally we are

01:34:12.960 --> 01:34:17.117
interested in being part of this not
just on the consumption side but also

01:34:17.200 --> 01:34:21.277
the production side of artificial
intelligence and certainly

01:34:21.360 --> 01:34:23.997
um through the provision of rare earths

01:34:24.080 --> 01:34:28.717
as well. But there is something that
goes beyond bilateral commercial or or

01:34:28.800 --> 01:34:33.037
national interests and that
is the interests of humanity

01:34:33.120 --> 01:34:35.197
beyond our own time beyond our own

01:34:35.280 --> 01:34:38.557
generation generation and the time that

01:34:38.640 --> 01:34:40.317
is allotted to each and every one of us.

01:34:40.400 --> 01:34:42.717
We are here now and we need to ask

01:34:42.800 --> 01:34:48.120
ourselves this differentiated question

01:34:48.720 --> 01:34:51.597
right now.

01:34:51.680 --> 01:34:55.357
We are setting this precedent which in

01:34:55.440 --> 01:34:58.317
the view of a country such as Colombia

01:34:58.400 --> 01:35:00.557
seems to be delicate. We regulated

01:35:00.640 --> 01:35:02.717
nuclear weapons. We regulate the

01:35:02.800 --> 01:35:05.757
manufacturing of tires. We even regulate

01:35:05.840 --> 01:35:10.637
the provision of homes. Anything that
could jeopardize the lives of the people

01:35:10.720 --> 01:35:15.437
that use those things is regulated. So
why then would we not regulate something

01:35:15.520 --> 01:35:17.357
which soon we are not even going to

01:35:17.440 --> 01:35:22.397
understand? That makes no sense.

01:35:22.480 --> 01:35:25.677
Super intelligence is not just the

01:35:25.760 --> 01:35:27.277
ability to achieve that through

01:35:27.360 --> 01:35:30.317
scientific transformation but also the

01:35:30.400 --> 01:35:33.757
ability to have that to know when we are

01:35:33.840 --> 01:35:38.317
no longer on the same level.
Thank you very much president.

01:35:38.400 --> 01:35:40.637
>> I thank the permanent representative of

01:35:40.720 --> 01:35:43.277
Colombia for the statement. I will now

01:35:43.360 --> 01:35:48.077
give the floor to her excellency, Miss
TZ Wagner, Minister of Foreign Affairs,

01:35:48.160 --> 01:35:52.157
International Cooperation, and Franco
of the Democratic Republic of the Congo.

01:35:52.240 --> 01:35:53.597
Thank you very much, Mr. President.

01:35:53.680 --> 01:35:56.717
Excellencies, ladies and gentlemen, the
Democratic Republic of the Congo. Thanks

01:35:56.800 --> 01:36:00.877
Frank thanks France as well as his
excellency Mr. Jean Noel Bahu, Minister

01:36:00.960 --> 01:36:04.157
for Europe and Foreign Affairs
for having convened and shared this

01:36:04.240 --> 01:36:06.877
important meeting. I also thank the

01:36:06.960 --> 01:36:10.157
briefers for their contributions to our

01:36:10.240 --> 01:36:12.397
collective reflection. Our debate takes

01:36:12.480 --> 01:36:15.037
place at a pivotal moment. Artificial

01:36:15.120 --> 01:36:19.437
intelligence is no longer merely
a matter of scientific progress or

01:36:19.520 --> 01:36:23.757
economic competitiveness. Because of the
speed of its developments, the scope of

01:36:23.840 --> 01:36:27.677
its applications, and its growing
integration into critical infrastructure

01:36:27.760 --> 01:36:30.637
and decision-making systems, it now

01:36:30.720 --> 01:36:33.757
raises questions that directly affect

01:36:33.840 --> 01:36:36.557
international peace and security.

01:36:36.640 --> 01:36:40.157
And this reality imposes
a responsibility upon us. the

01:36:40.240 --> 01:36:42.077
responsibility to anticipate risks

01:36:42.160 --> 01:36:47.197
without hindering innovation to preserve
the benefits of technological progress

01:36:47.280 --> 01:36:50.957
while preventing it from becoming
an additional factor of instability,

01:36:51.040 --> 01:36:53.917
vulnerability and conflict. The United

01:36:54.000 --> 01:36:57.997
Nations have begun laying
the foundations for this reflection. the

01:36:58.080 --> 01:37:02.397
work undertaken by the general
assembly on safe secure and trustworth

01:37:02.480 --> 01:37:04.397
trustworthy artificial intelligence the

01:37:04.480 --> 01:37:06.877
establishment of the independent

01:37:06.960 --> 01:37:11.757
international scientific panel on AI and
the global dialogue on AI governance as

01:37:11.840 --> 01:37:15.437
well as the security council's work
on the implications of scientific and

01:37:15.520 --> 01:37:19.517
technological advances for international
peace and security. All of these

01:37:19.600 --> 01:37:23.917
constitute important milestones.
It is now our responsibility to go further

01:37:24.000 --> 01:37:28.557
from anticipation to prevention and from
prevention to international cooperation

01:37:28.640 --> 01:37:32.077
that is to measure it with the
transformations underway. Presidents,

01:37:32.160 --> 01:37:34.797
the risks we face are numerous.

01:37:34.880 --> 01:37:38.957
Increasingly capable and autonomous
systems can create new vulnerabilities

01:37:39.040 --> 01:37:41.357
for critical infrastructure and

01:37:41.440 --> 01:37:44.157
decision-making processes. Their use in

01:37:44.240 --> 01:37:46.877
military and security contexts raises as

01:37:46.960 --> 01:37:50.157
fundamental questions of accountability,

01:37:50.240 --> 01:37:53.757
predictability, and human control. We

01:37:53.840 --> 01:37:56.877
must also remain vigilant regarding the

01:37:56.960 --> 01:37:59.437
unforeseen behavior of certain systems,

01:37:59.520 --> 01:38:01.917
the possibility of circumventing

01:38:02.000 --> 01:38:06.477
security safeguards, and the growing
difficulties states may encounter in

01:38:06.560 --> 01:38:10.557
understanding, assessing,
and regulating technologies

01:38:10.640 --> 01:38:12.877
whose capabilities are evolving at an

01:38:12.960 --> 01:38:15.997
unprecedented pace. These issues take on

01:38:16.080 --> 01:38:20.717
a particularly tangible dimension
in conflict situations. In the Democratic

01:38:20.800 --> 01:38:24.077
Republic of the Congo, as in other
theaters of crisis, surveillance

01:38:24.160 --> 01:38:27.997
technologies and drones already play
a growing role in the operational

01:38:28.080 --> 01:38:32.157
environment.
These technologies can contribute

01:38:32.240 --> 01:38:36.317
to reconnaissance, observation,
and ceasefire monitoring. They can also be

01:38:36.400 --> 01:38:38.877
diverted for military reconnaissance,

01:38:38.960 --> 01:38:41.037
coordination, disinformation, or

01:38:41.120 --> 01:38:43.917
propaganda purposes. This ambivalis

01:38:44.000 --> 01:38:45.997
reminds us of an essential truth. No

01:38:46.080 --> 01:38:48.637
technology is in itself an instrument of

01:38:48.720 --> 01:38:53.517
peace or conflict. It is the framework
within which it is designed, deployed

01:38:53.600 --> 01:38:55.517
and governed that largely determines its

01:38:55.600 --> 01:38:58.237
impact. That is why responsibility must

01:38:58.320 --> 01:39:00.557
remain at the heart of our reflection.

01:39:00.640 --> 01:39:03.437
States have a responsibility to

01:39:03.520 --> 01:39:08.157
establish appropriate legal frameworks,
ensure effective oversight, and prevent

01:39:08.240 --> 01:39:09.917
risks that could affect international

01:39:10.000 --> 01:39:12.317
peace and security.

01:39:12.400 --> 01:39:15.517
Technology companies must also

01:39:15.600 --> 01:39:19.517
fully assume their share of responsibility
particularly through risk

01:39:19.600 --> 01:39:24.157
assessments prior to deployment,
incident reporting, traceability,

01:39:24.240 --> 01:39:28.797
transparency and maintaining
appropriate human

01:39:28.880 --> 01:39:33.277
oversight. The scientific community
and civil society must likewise be able to

01:39:33.360 --> 01:39:36.557
contribute their independent expertise
and participate in the critical

01:39:36.640 --> 01:39:39.517
assessment of risks. But national

01:39:39.600 --> 01:39:44.237
responses, as necessary as they may
be, will not suffice. Technological

01:39:44.320 --> 01:39:48.477
interdependence calls for shared
responsibility. It requires

01:39:48.560 --> 01:39:52.157
closer international cooperation
based in particular on the exchange of

01:39:52.240 --> 01:39:55.277
information concerning serious incidents

01:39:55.360 --> 01:39:59.437
and identified vulnerabilities,
cooperation and evaluation and safety

01:39:59.520 --> 01:40:04.157
testing, as well as the establishment
of communication channels among governance

01:40:04.240 --> 01:40:08.077
government scientists and technology
stakeholders. President, for the

01:40:08.160 --> 01:40:09.597
Democratic Republic of the Congo, this

01:40:09.680 --> 01:40:12.317
cooperation must also meet a fundamental

01:40:12.400 --> 01:40:15.117
requirement, that of inclusion.

01:40:15.200 --> 01:40:19.837
Developing countries cannot be
reduced to the role of mere users of

01:40:19.920 --> 01:40:24.797
technologies designed elsewhere,
nor should be the mere recipients of

01:40:24.880 --> 01:40:26.397
standards and norms that they did not

01:40:26.480 --> 01:40:29.917
adequately participate in shaping.

01:40:30.000 --> 01:40:34.237
They must possess the scientific,
technical and institutional capacities

01:40:34.320 --> 01:40:37.117
that enable them to assess the
technologies to which their populations

01:40:37.200 --> 01:40:41.517
and economies are exposed to participate
in their governance and above all to

01:40:41.600 --> 01:40:43.917
contribute to innovation themselves.

01:40:44.000 --> 01:40:46.477
This requirement of inclusion leads us

01:40:46.560 --> 01:40:48.317
to consider artificial intelligence

01:40:48.400 --> 01:40:51.757
across its entire value chain.

01:40:51.840 --> 01:40:55.117
For despite its seemingly immaterial

01:40:55.200 --> 01:40:59.357
nature, the AI revolution rests
upon profoundly material realities,

01:40:59.440 --> 01:41:01.917
infrastructure, energy, industrial

01:41:02.000 --> 01:41:07.197
capabilities, and strategic natural
resources. This dimension is of

01:41:07.280 --> 01:41:11.517
particular importance to the Democratic
Republic of the Congo. During its

01:41:11.600 --> 01:41:13.357
presidency of the security council last

01:41:13.440 --> 01:41:16.157
July, my country invited the council to

01:41:16.240 --> 01:41:19.597
reflect on how responsible, transparent

01:41:19.680 --> 01:41:23.597
and equitable governance of natural
resources can contribute to conflict

01:41:23.680 --> 01:41:25.197
prevention,

01:41:25.280 --> 01:41:27.677
supply chain security

01:41:27.760 --> 01:41:32.477
and the emergence of shared prosperity.

01:41:32.560 --> 01:41:35.357
The AI revolution gives this reflection

01:41:35.440 --> 01:41:37.357
additional relevance. We must ensure

01:41:37.440 --> 01:41:39.517
that technological transformations of

01:41:39.600 --> 01:41:42.397
the 21st century do not reproduce the

01:41:42.480 --> 01:41:45.757
economic imbalances of previous decades.

01:41:45.840 --> 01:41:49.917
Countries supplying
essential resources to the

01:41:50.000 --> 01:41:52.557
global economy while processing

01:41:52.640 --> 01:41:54.637
technology and most of the added value

01:41:54.720 --> 01:41:57.197
remain concentrated elsewhere. The

01:41:57.280 --> 01:42:01.277
security of supply chains cannot be

01:42:01.360 --> 01:42:03.357
separated from the prosperity of the

01:42:03.440 --> 01:42:06.717
populations at their source. That is why

01:42:06.800 --> 01:42:09.437
AI governance must also be approached

01:42:09.520 --> 01:42:12.637
from a logic of shared responsibility.

01:42:12.720 --> 01:42:17.037
Responsibility in the design of systems,
responsibility in their use and

01:42:17.120 --> 01:42:22.157
responsibility within the value chains
that make their development possible and

01:42:22.240 --> 01:42:26.157
also responsibility in the distribution
of the benefits of technological

01:42:26.240 --> 01:42:29.917
progress.
For the Democratic Republic of the

01:42:30.000 --> 01:42:34.717
Congo, the objective
is clear. The resources

01:42:34.800 --> 01:42:37.917
that help make the digital revolution
possible must also contribute to the

01:42:38.000 --> 01:42:41.357
development, economic transformation,
industrialization and strengthening of

01:42:41.440 --> 01:42:45.677
the technological capabilities of those
countries that produce them. This is not

01:42:45.760 --> 01:42:47.837
merely a matter of development. It is

01:42:47.920 --> 01:42:52.717
also a matter of peace. When value
chains are opaque and benefits are

01:42:52.800 --> 01:42:56.797
distributed unequally
and when producing communities

01:42:56.880 --> 01:43:00.557
remain excluded from the prosperity
generated by their own resources,

01:43:00.640 --> 01:43:03.997
economic vulnerabilities can fuel tensions

01:43:04.080 --> 01:43:06.557
and conflict economies.

01:43:06.640 --> 01:43:08.637
Conversely, when resources are governed

01:43:08.720 --> 01:43:11.277
responsibly, when local processing

01:43:11.360 --> 01:43:13.917
creates jobs and opportunities,

01:43:14.000 --> 01:43:17.197
and when populations genuinely share in

01:43:17.280 --> 01:43:21.437
the benefits of growth, supply chains
can not only become more secure and

01:43:21.520 --> 01:43:25.677
resilient, but also chains
of shared prosperity.

01:43:25.760 --> 01:43:27.757
Mr. President, it is precisely here that

01:43:27.840 --> 01:43:30.957
the security council can bring added

01:43:31.040 --> 01:43:35.357
value. It is not the council's role
to replace the appropriate bodies

01:43:35.440 --> 01:43:37.197
responsible for developing the full

01:43:37.280 --> 01:43:39.837
range of technical rules relating to AI,

01:43:39.920 --> 01:43:42.877
but it is fully within the council's

01:43:42.960 --> 01:43:47.037
remit to anticipate the risks that may
affect international peace and security

01:43:47.120 --> 01:43:49.117
to consider how new technologies

01:43:49.200 --> 01:43:52.477
transform conflict dynamics, to protect

01:43:52.560 --> 01:43:56.557
civilians, to encourage confidence
building measures, and to ensure that

01:43:56.640 --> 01:44:00.557
technological transformations
do not create new vulnerabilities

01:44:00.640 --> 01:44:03.277
that could fuel and stability.

01:44:03.360 --> 01:44:07.517
Our objective must therefore not

01:44:07.600 --> 01:44:10.957
be to choose between innovation
and security any more than it should be to

01:44:11.040 --> 01:44:14.877
choose between natural resources
and development. Our responsibility is to

01:44:14.960 --> 01:44:17.917
govern progress so that it serves

01:44:18.000 --> 01:44:21.997
humanity. As July under
the presidency of the DRC,

01:44:22.080 --> 01:44:25.517
our council reflected on how to transform

01:44:25.600 --> 01:44:29.757
natural resources which are too often
associated with conflict economies into

01:44:29.840 --> 01:44:31.357
instruments of peace, cooperation, and

01:44:31.440 --> 01:44:33.837
shared prosperity.

01:44:33.920 --> 01:44:38.717
Today, the AI revolution in another form

01:44:38.800 --> 01:44:41.517
puts us in front of a similar question.

01:44:41.600 --> 01:44:45.117
How can we ensure that technological
progress reduces divisions rather than

01:44:45.200 --> 01:44:46.797
deepening them? How can we ensure that

01:44:46.880 --> 01:44:49.357
it strengthens states capacities rather

01:44:49.440 --> 01:44:52.477
than creating new dependencies?

01:44:52.560 --> 01:44:56.397
And finally, how can we ensure
that value chains, the value chains

01:44:56.480 --> 01:44:59.677
supporting this revolution contribute
to the peace and prosperity of the

01:44:59.760 --> 01:45:04.077
societies that make them possible
for the Democratic Republic of the Congo?

01:45:04.160 --> 01:45:06.317
The answer must rest on three

01:45:06.400 --> 01:45:08.077
principles: shared responsibility,

01:45:08.160 --> 01:45:12.397
inclusion, and shared prosperity.

01:45:12.480 --> 01:45:15.997
As the capabilities of artificial

01:45:16.080 --> 01:45:19.917
intelligence advance, our collective
capacity to govern them must advance

01:45:20.000 --> 01:45:24.157
alongside them. Because whether
the subject is natural resources or

01:45:24.240 --> 01:45:27.517
artificial intelligence, the fundamental
question remains the same. It is not

01:45:27.600 --> 01:45:31.837
only about what our resources
and technologies enable us to do, but also

01:45:31.920 --> 01:45:33.917
about what we collectively decide to do

01:45:34.000 --> 01:45:36.717
with them and that is where our common

01:45:36.800 --> 01:45:39.037
responsibility begins. Thank you. I

01:45:39.120 --> 01:45:44.157
thank her excellency Misswbo
Wagner for the statement and I

01:45:44.240 --> 01:45:49.400
now give the floor to the permanent
representative of China.

01:45:52.160 --> 01:45:55.837
>> Thank you president. I thank France
for convening this meeting and thank the

01:45:55.920 --> 01:45:58.157
briefers for their presentations.

01:45:58.240 --> 01:46:02.557
Artificial intelligence or AI is an
important driver of the new round of

01:46:02.640 --> 01:46:06.637
technological revolution, industrial
transformation and as such brings to

01:46:06.720 --> 01:46:09.917
humanity immense convenience
and opportunities for their life and

01:46:10.000 --> 01:46:13.437
productive activities while also
creating new issues and challenges for

01:46:13.520 --> 01:46:16.957
international peace and security.
It is incumbent upon the international

01:46:17.040 --> 01:46:20.877
community to work more closely together,
attach equal importance to development

01:46:20.960 --> 01:46:23.917
and security and steer AI development

01:46:24.000 --> 01:46:28.717
for the positive for good and for
humanity thereby contributing to world

01:46:28.800 --> 01:46:32.397
peace, stability, development
and prosperity. I wish to underscore the

01:46:32.480 --> 01:46:36.717
following points. The first point is about
respecting digital sovereignty and

01:46:36.800 --> 01:46:39.117
advocating openness for mutual benefit.

01:46:39.200 --> 01:46:41.597
As AI technology is deeply embedded in

01:46:41.680 --> 01:46:45.357
global production and supply chains and

01:46:45.440 --> 01:46:48.317
innovation networks,

01:46:48.400 --> 01:46:52.957
all countries have the right
to independently choose AI technologies,

01:46:53.040 --> 01:46:56.717
products and services following
the dictates of their actual needs and

01:46:56.800 --> 01:46:59.117
market principles.

01:46:59.200 --> 01:47:01.997
Banding together into a petty us versus

01:47:02.080 --> 01:47:04.717
them clique in the tech sector and

01:47:04.800 --> 01:47:09.437
forcing other countries to take sides
is tantamount to politicizing and

01:47:09.520 --> 01:47:11.677
instrumentalizing tech issues in the

01:47:11.760 --> 01:47:14.557
service of certain major powers attempt

01:47:14.640 --> 01:47:16.957
to maintain their technological edge

01:47:17.040 --> 01:47:19.277
through monopoly in their quest for

01:47:19.360 --> 01:47:21.917
digital hedgeimonyy. This not only

01:47:22.000 --> 01:47:26.317
hampers global tech progress to the
detriment of the development rights and

01:47:26.400 --> 01:47:30.557
interests of all countries, especially
smaller nations, but also increases the

01:47:30.640 --> 01:47:34.557
risk of fragmentation of technical
standards and incompatibility between

01:47:34.640 --> 01:47:39.277
systems and thus putting global security
in serious jeopardy. We call on all

01:47:39.360 --> 01:47:41.277
countries to enhance exchanges and

01:47:41.360 --> 01:47:44.397
cooperation, promote open

01:47:44.480 --> 01:47:48.477
source and open ecosystems, and work
together for global scientific and

01:47:48.560 --> 01:47:52.477
technological advancement. My second
point is about strengthening risk

01:47:52.560 --> 01:47:55.117
awareness and ensuring that AI is secure

01:47:55.200 --> 01:47:57.917
and controllable. All countries should

01:47:58.000 --> 01:48:01.437
uphold the people centered philosophy.

01:48:01.520 --> 01:48:05.997
Making sure AI research, development
and application are in line with the

01:48:06.080 --> 01:48:10.157
purposes and principles of the UN
charter and applicable international law

01:48:10.240 --> 01:48:12.877
as well as the common values of humanity

01:48:12.960 --> 01:48:16.077
and social ethics. In response to the

01:48:16.160 --> 01:48:18.957
inherent and secondary risks posed by

01:48:19.040 --> 01:48:23.837
AI, it is necessary to continuously
improve improve regulatory frameworks,

01:48:23.920 --> 01:48:25.677
emergency response and crossber

01:48:25.760 --> 01:48:28.317
cooperation mechanisms and resolutely

01:48:28.400 --> 01:48:31.277
combat the malicious abuse of AI by

01:48:31.360 --> 01:48:34.317
terrorist organizations, extremist

01:48:34.400 --> 01:48:37.117
forces and transnational criminal

01:48:37.200 --> 01:48:40.477
groups. Relevant businesses, especially

01:48:40.560 --> 01:48:43.357
tech giants, should fulfill their social

01:48:43.440 --> 01:48:45.917
responsibilities, counter disinformation

01:48:46.000 --> 01:48:48.157
and hate speech, and subject themselves

01:48:48.240 --> 01:48:50.877
to regulation by law and oversight by

01:48:50.960 --> 01:48:55.757
the public. Regarding AI, R&D and
application in the military domain, all

01:48:55.840 --> 01:48:57.837
countries should take a prudent and

01:48:57.920 --> 01:49:00.477
responsible approach. avoid strategic

01:49:00.560 --> 01:49:03.997
miscalculations and an arms race. Oppose

01:49:04.080 --> 01:49:08.717
the use of related weapons
for indiscriminate attacks or deliberate

01:49:08.800 --> 01:49:10.957
strikes against civilians and civilian

01:49:11.040 --> 01:49:13.357
facilities and make sure that the

01:49:13.440 --> 01:49:16.237
ultimate decision making power stays

01:49:16.320 --> 01:49:19.117
with humans. The third point is about

01:49:19.200 --> 01:49:21.597
upholding AI for good and strengthening

01:49:21.680 --> 01:49:24.717
technological empowerment. AI has

01:49:24.800 --> 01:49:28.877
unlocked many possibilities
for addressing global challenges such as

01:49:28.960 --> 01:49:32.957
those related to climate change,
food security, public health, energy

01:49:33.040 --> 01:49:35.757
transition, and natural disasters. The

01:49:35.840 --> 01:49:37.357
international community should seize

01:49:37.440 --> 01:49:40.237
this historic opportunity, support the

01:49:40.320 --> 01:49:43.677
empowerment of all industries through AI

01:49:43.760 --> 01:49:47.197
technologies and provide guidance for

01:49:47.280 --> 01:49:49.277
technological advances to better

01:49:49.360 --> 01:49:52.157
translate into improvements in public

01:49:52.240 --> 01:49:56.557
well-being. We welcome the United
Nations efforts to make the best use of

01:49:56.640 --> 01:50:00.237
existing resources and apply AI
technology to enhance the quality and

01:50:00.320 --> 01:50:02.237
efficiency of its work thereby providing

01:50:02.320 --> 01:50:04.877
more agile and responsive service to

01:50:04.960 --> 01:50:09.837
member states. UN field missions with
the consent of the host governments and

01:50:09.920 --> 01:50:13.757
in line with their respective mandates
may use AI technology properly to

01:50:13.840 --> 01:50:18.157
strengthen their situational awareness,
information analysis analysis, risk

01:50:18.240 --> 01:50:22.317
early warning and rapid response
capabilities. My fourth point is about

01:50:22.400 --> 01:50:25.117
promoting fairness and inclusivity and

01:50:25.200 --> 01:50:27.677
improving global governance. AI is a

01:50:27.760 --> 01:50:30.637
shared asset for all humanity and should

01:50:30.720 --> 01:50:33.037
not become rich countries and rich

01:50:33.120 --> 01:50:35.757
people's game. We must prevent the

01:50:35.840 --> 01:50:39.757
digital and AI divides from exacerbating

01:50:39.840 --> 01:50:42.717
the development divide. Increase support

01:50:42.800 --> 01:50:46.077
for the global south. help people of all

01:50:46.160 --> 01:50:50.877
countries peacefully partake of the
dividends of technological progress and

01:50:50.960 --> 01:50:52.557
make them more resilient against

01:50:52.640 --> 01:50:55.837
security risks. We must practice true

01:50:55.920 --> 01:51:01.037
multilateralism, support the UN as the
main forum of global AI governance and

01:51:01.120 --> 01:51:02.797
give developing countries more

01:51:02.880 --> 01:51:05.597
representation and a bigger voice. On

01:51:05.680 --> 01:51:08.077
the basis of respecting the diversity of

01:51:08.160 --> 01:51:10.717
civilizations and the uniqueness of each

01:51:10.800 --> 01:51:12.957
country's culture, we should enhance

01:51:13.040 --> 01:51:15.597
alignment and coordination on AI

01:51:15.680 --> 01:51:18.077
development strategies, governance rules

01:51:18.160 --> 01:51:21.117
and technical standards and form a

01:51:21.200 --> 01:51:22.797
consensusbased

01:51:22.880 --> 01:51:25.117
global governance framework at an early

01:51:25.200 --> 01:51:28.157
date so as to build institutional guard

01:51:28.240 --> 01:51:30.477
rails that preserve international peace

01:51:30.560 --> 01:51:33.357
and security. President China is an

01:51:33.440 --> 01:51:36.477
important practitioner in AI innovation

01:51:36.560 --> 01:51:40.477
and development and an active
contributor to global governance. Since

01:51:40.560 --> 01:51:43.357
President Xi Jinping launched the global

01:51:43.440 --> 01:51:46.317
AI governance initiative, China has

01:51:46.400 --> 01:51:49.597
helped the UN General Assembly adopt by

01:51:49.680 --> 01:51:52.877
consensus the resolution on enhancing

01:51:52.960 --> 01:51:54.797
international cooperation on capacity

01:51:54.880 --> 01:51:57.117
building of artificial intelligence and

01:51:57.200 --> 01:52:00.397
tabled the AI capacity building action

01:52:00.480 --> 01:52:03.997
plan for good and for all and the AI

01:52:04.080 --> 01:52:05.757
plus international cooperation

01:52:05.840 --> 01:52:08.397
initiative. In July this year, President

01:52:08.480 --> 01:52:11.117
Xi Jinping announced at the World AI

01:52:11.200 --> 01:52:13.277
Conference that in the next five years,

01:52:13.360 --> 01:52:16.477
China will provide 5,000 opportunities

01:52:16.560 --> 01:52:19.597
in AI training and seminar programs for

01:52:19.680 --> 01:52:22.797
developing countries.

01:52:22.880 --> 01:52:27.437
build international AI application
cooperation centers for ASEAN, the

01:52:27.520 --> 01:52:29.357
League of Arab States, the African

01:52:29.440 --> 01:52:33.117
Union, SELAC, the Shanghai Corporation

01:52:33.200 --> 01:52:35.757
Organization and BRICS, and advanced the

01:52:35.840 --> 01:52:38.077
application of the AI powered

01:52:38.160 --> 01:52:41.357
meteorological warning system named MASU

01:52:41.440 --> 01:52:45.117
in 30 countries proposed and championed

01:52:45.200 --> 01:52:47.837
by China and in response to the call of

01:52:47.920 --> 01:52:50.637
our time, the World Artificial

01:52:50.720 --> 01:52:55.597
Intelligence Cooperation Organization
ation or Wiko for short was founded in

01:52:55.680 --> 01:52:58.877
the city of Shanghai. It is the
world's first ever intergovernmental

01:52:58.960 --> 01:53:02.637
international organization dedicated
to AI. We welcome more countries to join

01:53:02.720 --> 01:53:06.557
this open platform and look forward to
working with all nations to help build a

01:53:06.640 --> 01:53:11.117
just and equitable system for global
AI governance. Together we can and shall

01:53:11.200 --> 01:53:12.877
create brighter future for humanity.

01:53:12.960 --> 01:53:16.877
Thank you, President. >> I thank
the permanent representative of

01:53:16.960 --> 01:53:19.037
China for this statement. I now give the

01:53:19.120 --> 01:53:24.120
floor to the representative
of the Russian Federation.

01:53:26.560 --> 01:53:30.637
Mr. President, we thank
the briefers for their

01:53:30.720 --> 01:53:33.117
contributions to the discussion today.

01:53:33.200 --> 01:53:36.797
Artificial intelligence has become quite
a trendy topic and there is virtually no

01:53:36.880 --> 01:53:40.477
one in the world who is indifferent to it.
It is now being discussed in a

01:53:40.560 --> 01:53:43.357
variety of political and expert fora and

01:53:43.440 --> 01:53:45.677
the security council is no exception.

01:53:45.760 --> 01:53:47.677
However, this principal organ of the

01:53:47.760 --> 01:53:51.037
United Nations has clear limitations.

01:53:51.120 --> 01:53:53.997
First of all, it is hard to see how the

01:53:54.080 --> 01:53:55.917
generic topic of artificial intelligence

01:53:56.000 --> 01:53:59.677
is related to the council's mandate as

01:53:59.760 --> 01:54:03.437
clearly defined in the UN charter
for the maintenance of international peace

01:54:03.520 --> 01:54:05.277
and security. Secondly, and equally

01:54:05.360 --> 01:54:07.997
importantly, the countries of Africa,

01:54:08.080 --> 01:54:10.797
Asia, and Latin America are under

01:54:10.880 --> 01:54:13.757
represented in the security council. The

01:54:13.840 --> 01:54:16.237
voice of the global south is unlikely to

01:54:16.320 --> 01:54:18.557
be heard in full. Accordingly, there is

01:54:18.640 --> 01:54:23.357
a high risk that the dominant
group of Western states in the

01:54:23.440 --> 01:54:27.277
council, which is interested
in preserving and consolidating its

01:54:27.360 --> 01:54:29.837
technological leadership in this area,

01:54:29.920 --> 01:54:32.717
will seek to impose

01:54:32.800 --> 01:54:37.517
its narrow self-serving approach on the
entire international community to the

01:54:37.600 --> 01:54:39.517
detriment of the prerogative of

01:54:39.600 --> 01:54:42.157
inclusive specialized mechanisms engaged

01:54:42.240 --> 01:54:46.077
in practical work in this field.

01:54:46.160 --> 01:54:50.840
And regrettably,
this is already happening.

01:54:51.040 --> 01:54:53.757
One striking example is the holding of

01:54:53.840 --> 01:54:56.717
closed door summits on the so-called

01:54:56.800 --> 01:55:00.237
responsible military use of

01:55:00.320 --> 01:55:03.757
AI to which most developing countries are

01:55:03.840 --> 01:55:08.877
not invited. Finally,

01:55:08.960 --> 01:55:10.717
issues relating to both the civilian and

01:55:10.800 --> 01:55:14.317
military applications of AI technologies

01:55:14.400 --> 01:55:17.677
are complex and multi-dimensional

01:55:17.760 --> 01:55:21.757
and that is precisely why these issues

01:55:21.840 --> 01:55:23.997
should be discussed in relevant

01:55:24.080 --> 01:55:26.797
specialized fora. For instance,

01:55:26.880 --> 01:55:32.200
discussions of AI enabled cyber attacks

01:55:32.320 --> 01:55:36.317
would be more appropriate
within the global

01:55:36.400 --> 01:55:38.797
mechanism established as the successor

01:55:38.880 --> 01:55:41.917
to the UN open-ended working group on

01:55:42.000 --> 01:55:46.920
security of and in the use of ICTs.

01:55:48.240 --> 01:55:49.997
And for the consideration of the

01:55:50.080 --> 01:55:53.677
military aspects of AI, there is the

01:55:53.760 --> 01:55:55.917
group of governmental experts on lethal

01:55:56.000 --> 01:56:00.717
autonomous weapons systems which
operates under the framework of the

01:56:00.800 --> 01:56:05.640
convention on certain
conventional weapons.

01:56:06.160 --> 01:56:08.157
Would like to draw colleagues attention

01:56:08.240 --> 01:56:12.717
to the fact that the GGE in Geneva

01:56:12.800 --> 01:56:17.357
adopted a substantive consensus
report just three weeks ago.

01:56:17.440 --> 01:56:21.357
This outcome serves as a
practical confirmation of

01:56:21.440 --> 01:56:23.837
the effectiveness and relevance of that

01:56:23.920 --> 01:56:27.037
forum. Most importantly, the report

01:56:27.120 --> 01:56:31.197
contains a working definition
of lethal autonomous weapons systems,

01:56:31.280 --> 01:56:34.557
common understandings regarding
the applicability of the norms of

01:56:34.640 --> 01:56:36.237
international law to such systems,

01:56:36.320 --> 01:56:37.917
including international humanitarian

01:56:38.000 --> 01:56:42.477
law, as well as parameters

01:56:42.560 --> 01:56:45.357
for ensuring human control in order to

01:56:45.440 --> 01:56:52.280
mitigate risks. associated
with the military use of AI.

01:56:54.000 --> 01:56:58.237
We do not believe it would be advisable
to undermine the constructive work

01:56:58.320 --> 01:57:02.317
there's already been accomplished
in Geneva. For our part, we are ready to

01:57:02.400 --> 01:57:04.237
continue participating actively in the

01:57:04.320 --> 01:57:06.637
group's work with the view to further

01:57:06.720 --> 01:57:10.877
aligning common approaches in this area.

01:57:10.960 --> 01:57:14.280
Mr. President,

01:57:14.400 --> 01:57:17.357
we have carefully studied the concept

01:57:17.440 --> 01:57:21.997
note for today's meeting
circulated by the French presidency.

01:57:22.080 --> 01:57:23.837
We believe it is counterproductive to

01:57:23.920 --> 01:57:25.837
consider the use of AI technologies

01:57:25.920 --> 01:57:29.197
solely through the lens of risk.

01:57:29.280 --> 01:57:32.157
After all, every coin has two sides. The

01:57:32.240 --> 01:57:34.637
advantages of AI include reliability,

01:57:34.720 --> 01:57:38.317
security, and above all speed,

01:57:38.400 --> 01:57:41.357
including in decision-m.

01:57:41.440 --> 01:57:44.957
AI allows to carry out an effective

01:57:45.040 --> 01:57:49.437
analysis of large data sets.

01:57:49.520 --> 01:57:53.277
It can help to identify threats, enhance

01:57:53.360 --> 01:57:55.997
situational awareness, and improve

01:57:56.080 --> 01:58:00.957
precision in the conduct of operations
as well as reduce risks to the civilian

01:58:01.040 --> 01:58:03.997
population and strengthen
the protection of

01:58:04.080 --> 01:58:05.917
civilian objects.

01:58:06.000 --> 01:58:10.397
At the same time, of course,

01:58:10.480 --> 01:58:11.917
it must be borne in mind that this

01:58:12.000 --> 01:58:14.397
technology also entails risks.

01:58:14.480 --> 01:58:17.677
minimizing them is

01:58:17.760 --> 01:58:19.517
first and foremost a matter of ensuring

01:58:19.600 --> 01:58:25.160
human control over the functioning
of AI enabled technologies.

01:58:25.600 --> 01:58:27.917
The specific forms of such controls are

01:58:28.000 --> 01:58:34.120
determined by states in practice
at the national level.

01:58:34.480 --> 01:58:36.637
Some possible means of ensuring human

01:58:36.720 --> 01:58:40.397
control include, for example,

01:58:40.480 --> 01:58:44.877
enhancing system
reliability and resiliency,

01:58:44.960 --> 01:58:48.077
limiting the types of targets, restricting

01:58:48.160 --> 01:58:50.637
the duration of operation, geographical

01:58:50.720 --> 01:58:54.557
scope and scale of use,

01:58:54.640 --> 01:58:57.517
and timely intervention and deactivation

01:58:57.600 --> 01:58:59.437
of the system.

01:58:59.520 --> 01:59:02.317
In addition,

01:59:02.400 --> 01:59:04.877
steps must also undertake measures to

01:59:04.960 --> 01:59:07.357
detect and minimize errors in algorithms

01:59:07.440 --> 01:59:11.597
and data sets as well as adopt

01:59:11.680 --> 01:59:16.477
appropriate measures to prevent and
suppress violations of international law

01:59:16.560 --> 01:59:21.277
in the context of the development
and use of relevant technologies. Such

01:59:21.360 --> 01:59:23.357
measures

01:59:23.440 --> 01:59:25.997
could include effective life cycle

01:59:26.080 --> 01:59:28.717
management of weapons systems and

01:59:28.800 --> 01:59:31.597
military equipment,

01:59:31.680 --> 01:59:33.837
establishing readiness criteria, and

01:59:33.920 --> 01:59:39.757
ensuring maximum protection
from unauthorized access.

01:59:39.840 --> 01:59:45.037
We also see added value
in exchanging views

01:59:45.120 --> 01:59:47.517
and developing specific recommendations

01:59:47.600 --> 01:59:50.717
and guidelines. However, it is important

01:59:50.800 --> 01:59:52.797
that this be done exclusively in

01:59:52.880 --> 01:59:56.397
relevant specialized bodies.

01:59:56.480 --> 02:00:00.077
Russia does not object to a multilateral

02:00:00.160 --> 02:00:02.797
dialogue on the applications of AI in

02:00:02.880 --> 02:00:05.437
principle and this includes the military

02:00:05.520 --> 02:00:09.837
applications of AI as well
as considering the implications of the use

02:00:09.920 --> 02:00:11.437
of such tools for international peace

02:00:11.520 --> 02:00:13.997
and security. We are convinced however

02:00:14.080 --> 02:00:17.037
that such a dialogue must be based on

02:00:17.120 --> 02:00:20.317
consensus and inclusiveness

02:00:20.400 --> 02:00:23.197
and be intergovernmental in nature.

02:00:23.280 --> 02:00:28.077
These fundamental principles are
intended to guarantee guarantee that

02:00:28.160 --> 02:00:29.917
the interests of every member state are

02:00:30.000 --> 02:00:32.797
taken into account and to ensure that

02:00:32.880 --> 02:00:35.277
the decisions adopted are effective.

02:00:35.360 --> 02:00:37.277
Moreover, in exchanging views on this

02:00:37.360 --> 02:00:39.597
subject,

02:00:39.680 --> 02:00:42.877
a particular focus must be placed on

02:00:42.960 --> 02:00:46.157
ensuring nondiscriminatory and equitable

02:00:46.240 --> 02:00:51.037
access for all states, particularly
the countries of the global south, to AI

02:00:51.120 --> 02:00:55.117
technologies.
Thank you for your attention.

02:00:55.200 --> 02:00:57.677
>> I thank the representative of the
Russian Federation for his statement. I

02:00:57.760 --> 02:01:01.917
now give the floor to the permanent
representative of Liberia.

02:01:02.000 --> 02:01:04.637
>> Thank you, Mr. President. Liberia joins

02:01:04.720 --> 02:01:07.757
in thanking the presidency of France for

02:01:07.840 --> 02:01:09.437
convening

02:01:09.520 --> 02:01:13.357
this critical debate

02:01:13.440 --> 02:01:16.797
and we thank the briefers for sharing

02:01:16.880 --> 02:01:20.237
their qualified perspectives.

02:01:20.320 --> 02:01:22.797
We also join in welcoming the high level

02:01:22.880 --> 02:01:27.037
dignitaries participating.

02:01:27.120 --> 02:01:29.437
Excellencies,

02:01:29.520 --> 02:01:33.197
history teaches us that humanity's

02:01:33.280 --> 02:01:35.677
greatest transformations

02:01:35.760 --> 02:01:38.717
are never defined by technological

02:01:38.800 --> 02:01:41.437
breakthroughs alone.

02:01:41.520 --> 02:01:45.597
Every fundamental leap in capability

02:01:45.680 --> 02:01:49.117
has demanded an equal leap in

02:01:49.200 --> 02:01:50.637
governance,

02:01:50.720 --> 02:01:53.757
ethical frameworks and multilateral

02:01:53.840 --> 02:01:57.277
rules to ensure that power remains

02:01:57.360 --> 02:02:01.197
titted to the common good.

02:02:01.280 --> 02:02:03.837
Artificial intelligence

02:02:03.920 --> 02:02:07.757
presents the defining test of that

02:02:07.840 --> 02:02:10.557
timeless principle.

02:02:10.640 --> 02:02:13.997
Its promise to uplift societies

02:02:14.080 --> 02:02:17.117
is immense.

02:02:17.200 --> 02:02:22.157
So too is the existential peril should

02:02:22.240 --> 02:02:24.077
its trajectory

02:02:24.160 --> 02:02:27.597
strip human control, institutional

02:02:27.680 --> 02:02:34.520
accountability and ethical oversight.

02:02:36.240 --> 02:02:39.277
81 years ago,

02:02:39.360 --> 02:02:41.437
the architects of the United Nations

02:02:41.520 --> 02:02:45.917
Charter confronted a world fractured by

02:02:46.000 --> 02:02:52.920
unrestrained power. They recognized

02:02:53.360 --> 02:02:58.077
that sustainable peace cannot be secured

02:02:58.160 --> 02:03:02.397
merely by reacting to crisis after the

02:03:02.480 --> 02:03:05.677
smoke has cleared.

02:03:05.760 --> 02:03:09.037
It demands foresight, sustained

02:03:09.120 --> 02:03:11.597
multilateral cooperation

02:03:11.680 --> 02:03:15.357
and institutions capable of anticipating

02:03:15.440 --> 02:03:19.437
and neutralizing threats before they

02:03:19.520 --> 02:03:23.757
ignite global conflict.

02:03:23.840 --> 02:03:26.957
While the delegates at San Francisco in

02:03:27.040 --> 02:03:28.957
1945

02:03:29.040 --> 02:03:31.837
could not have envisioned neural

02:03:31.920 --> 02:03:36.477
networks or frontier computing models,

02:03:36.560 --> 02:03:39.837
any capability that reshapes the nature

02:03:39.920 --> 02:03:41.837
of warfare,

02:03:41.920 --> 02:03:45.997
sovereignty, and geopolitical stability

02:03:46.080 --> 02:03:49.837
falls squarely within this council's

02:03:49.920 --> 02:03:52.477
primary mandate to maintain

02:03:52.560 --> 02:03:57.197
international peace. peace and security.

02:03:57.280 --> 02:03:59.597
Mr. President,

02:03:59.680 --> 02:04:02.317
artificial intelligence

02:04:02.400 --> 02:04:05.917
is already reshaping critical

02:04:06.000 --> 02:04:07.837
infrastructure,

02:04:07.920 --> 02:04:12.957
global commerce, and social cohesion.

02:04:13.040 --> 02:04:15.837
Governed responsibly,

02:04:15.920 --> 02:04:19.037
these technologies can empower the

02:04:19.120 --> 02:04:23.277
United Nations, enhancing conflict,

02:04:23.360 --> 02:04:26.237
early warning mechanisms,

02:04:26.320 --> 02:04:29.917
optimizing humanitarian logistics,

02:04:30.000 --> 02:04:31.917
and strengthening peacekeeping

02:04:32.000 --> 02:04:34.077
operations.

02:04:34.160 --> 02:04:39.197
Yet we observe with grave concern the

02:04:39.280 --> 02:04:42.797
proliferation of AI enabled offensive

02:04:42.880 --> 02:04:44.557
capabilities,

02:04:44.640 --> 02:04:48.077
autonomous weapon systems and frontier

02:04:48.160 --> 02:04:52.237
models exhibiting unpredictable

02:04:52.320 --> 02:04:55.277
emergent behaviors.

02:04:55.360 --> 02:04:59.037
Rapid advances toward artificial general

02:04:59.120 --> 02:05:00.557
intelligence

02:05:00.640 --> 02:05:05.277
and highly autonomous systems underscore

02:05:05.360 --> 02:05:08.077
that frontier risks are no longer

02:05:08.160 --> 02:05:11.757
abstract theoretical possibilities.

02:05:11.840 --> 02:05:14.157
They are truly

02:05:14.240 --> 02:05:18.397
eminent policy realities.

02:05:18.480 --> 02:05:21.677
The international comm community cannot

02:05:21.760 --> 02:05:26.317
afford to wait for a systemic failure or

02:05:26.400 --> 02:05:29.197
catastrophic miscalculation

02:05:29.280 --> 02:05:31.677
before establishing collective

02:05:31.760 --> 02:05:33.357
safeguards.

02:05:33.440 --> 02:05:37.357
We must not allow the cost of today's

02:05:37.440 --> 02:05:40.477
hesitation to become tomorrow's

02:05:40.560 --> 02:05:45.197
catastrophe. Mr. President

02:05:45.280 --> 02:05:47.277
Labbera believes the international

02:05:47.360 --> 02:05:52.157
community must steer clear of two

02:05:52.240 --> 02:05:56.717
equally hazardous extremes.

02:05:56.800 --> 02:06:01.197
The first is suffocating overregulation

02:06:01.280 --> 02:06:04.557
that stifles progress particularly for

02:06:04.640 --> 02:06:07.597
developing nations that urgently need

02:06:07.680 --> 02:06:12.477
access to AI tools to bolster education.

02:06:12.560 --> 02:06:15.677
revolutionize health care delivery,

02:06:15.760 --> 02:06:18.077
modernize agriculture

02:06:18.160 --> 02:06:20.797
and accelerate the sustainable

02:06:20.880 --> 02:06:24.637
development goals. The second is

02:06:24.720 --> 02:06:27.597
dangerous complacency.

02:06:27.680 --> 02:06:30.637
The naive assumption that transformative

02:06:30.720 --> 02:06:36.157
technologies can safely self-regulate.

02:06:36.240 --> 02:06:40.157
History offers no precedent for such a

02:06:40.240 --> 02:06:41.917
faith.

02:06:42.000 --> 02:06:45.357
Without verifiable safeguards,

02:06:45.440 --> 02:06:49.117
transparency and clear liability,

02:06:49.200 --> 02:06:53.037
unregulated capabilities will generate

02:06:53.120 --> 02:06:54.557
vulnerabilities

02:06:54.640 --> 02:06:59.117
that respect no national border.

02:06:59.200 --> 02:07:03.757
Our path is not a binary choice between

02:07:03.840 --> 02:07:07.677
innovation and regulation.

02:07:07.760 --> 02:07:11.597
The imperative before us is to ensure

02:07:11.680 --> 02:07:14.237
that technological innovation and

02:07:14.320 --> 02:07:17.997
multilateral governance advance

02:07:18.080 --> 02:07:23.517
hand in hand. We must regulate with

02:07:23.600 --> 02:07:29.757
wisdom and innovate with responsibility.

02:07:29.840 --> 02:07:31.757
For Liberia,

02:07:31.840 --> 02:07:36.557
four operational pillars are paramount.

02:07:36.640 --> 02:07:38.557
One,

02:07:38.640 --> 02:07:42.157
safety by design

02:07:42.240 --> 02:07:45.357
responsibility must begin at the

02:07:45.440 --> 02:07:47.997
architecture phase.

02:07:48.080 --> 02:07:51.597
Rigorous risk mitigation, cyber security

02:07:51.680 --> 02:07:55.357
resilience and fail safe controls must

02:07:55.440 --> 02:07:59.117
be embedded across the entire life cycle

02:07:59.200 --> 02:08:02.637
from pre-training to public deployment

02:08:02.720 --> 02:08:05.117
rather than improvised

02:08:05.200 --> 02:08:07.917
post crisis.

02:08:08.000 --> 02:08:11.837
Two, upholding international law and

02:08:11.920 --> 02:08:14.157
human agency.

02:08:14.240 --> 02:08:16.717
Decisions involving the use of lethal

02:08:16.800 --> 02:08:19.917
force, the deprivation of human life,

02:08:20.000 --> 02:08:22.557
and the escalation of conflict must

02:08:22.640 --> 02:08:25.997
never be abdicated to autonomous

02:08:26.080 --> 02:08:29.917
algorithms. Meaningful human control and

02:08:30.000 --> 02:08:31.517
accountability

02:08:31.600 --> 02:08:34.637
remain the non-negotiable bedrock of

02:08:34.720 --> 02:08:37.917
international humanitarian law and the

02:08:38.000 --> 02:08:40.477
human dignity.

02:08:40.560 --> 02:08:43.357
Collaborative confidence building.

02:08:43.440 --> 02:08:45.517
Three,

02:08:45.600 --> 02:08:47.917
multilateral

02:08:48.000 --> 02:08:51.597
cooperation must foster trust without

02:08:51.680 --> 02:08:54.957
impeding scientific inquiry.

02:08:55.040 --> 02:08:57.757
Harmonized safety benchmarks,

02:08:57.840 --> 02:09:00.637
coordinated threat disclosures, and

02:09:00.720 --> 02:09:03.757
transparent red teameing protocols

02:09:03.840 --> 02:09:07.437
between states, industry leaders, and

02:09:07.520 --> 02:09:11.117
independent scientists are essential to

02:09:11.200 --> 02:09:14.557
mitigate strategic mistrust.

02:09:14.640 --> 02:09:18.477
And four, independent verifiable

02:09:18.560 --> 02:09:20.717
oversight.

02:09:20.800 --> 02:09:25.037
Norms without verification are

02:09:25.120 --> 02:09:27.117
insufficient.

02:09:27.200 --> 02:09:30.957
We require credible auditing frameworks,

02:09:31.040 --> 02:09:33.997
robust red teaming standards and

02:09:34.080 --> 02:09:36.637
independent technical evaluation to

02:09:36.720 --> 02:09:39.997
ensure that corporate and state

02:09:40.080 --> 02:09:43.757
commitments translate into measurable

02:09:43.840 --> 02:09:48.120
safety for mankind.

02:09:48.480 --> 02:09:51.677
Furthermore, Mr. President,

02:09:51.760 --> 02:09:55.117
global governance cannot become an

02:09:55.200 --> 02:09:57.517
exclusive enclave

02:09:57.600 --> 02:10:00.077
reserved for the few nations that

02:10:00.160 --> 02:10:04.477
control advanced compute and capital.

02:10:04.560 --> 02:10:08.157
Africa and the global south cannot be

02:10:08.240 --> 02:10:11.837
treated merely as data extractors or

02:10:11.920 --> 02:10:16.397
downstream end users. We must be equal

02:10:16.480 --> 02:10:19.677
co-architects in determining the

02:10:19.760 --> 02:10:21.197
standards,

02:10:21.280 --> 02:10:24.957
ethics and architectures of this

02:10:25.040 --> 02:10:29.320
technological era.

02:10:30.000 --> 02:10:33.037
In this regard, Labber welcomes the

02:10:33.120 --> 02:10:35.437
landmark initiatives advanced through

02:10:35.520 --> 02:10:38.717
the Global Digital Compact, notably the

02:10:38.800 --> 02:10:42.317
Global Dialogue on AI governance and the

02:10:42.400 --> 02:10:47.517
independent international
scientific panel on AI.

02:10:47.600 --> 02:10:49.037
However,

02:10:49.120 --> 02:10:51.037
complimentary efforts in the General

02:10:51.120 --> 02:10:54.797
Assembly do not absolve this council of

02:10:54.880 --> 02:10:57.757
its specific duty.

02:10:57.840 --> 02:11:00.477
When advanced computing intersects with

02:11:00.560 --> 02:11:03.517
cyber operations against civilian

02:11:03.600 --> 02:11:05.037
infrastructure,

02:11:05.120 --> 02:11:08.237
autonomous weaponized systems and

02:11:08.320 --> 02:11:12.157
geostrategic competition, it intersects

02:11:12.240 --> 02:11:15.677
directly with international security.

02:11:15.760 --> 02:11:19.917
Council engagement is not duplication.

02:11:20.000 --> 02:11:23.357
It is conflict prevention.

02:11:23.440 --> 02:11:24.877
Liberia

02:11:24.960 --> 02:11:28.717
therefore lends it strong support to the

02:11:28.800 --> 02:11:31.197
establishment of a dedicated security

02:11:31.280 --> 02:11:34.077
council informal working group on

02:11:34.160 --> 02:11:36.957
artificial intelligence and emerging

02:11:37.040 --> 02:11:40.557
technologies tasked with monitoring the

02:11:40.640 --> 02:11:43.677
crosscutting security implications of

02:11:43.760 --> 02:11:47.597
these capabilities across both thematic

02:11:47.680 --> 02:11:51.197
and country specific agendas. To

02:11:51.280 --> 02:11:54.557
conclude, Mr. President,

02:11:54.640 --> 02:11:57.837
let us act with the foresight and

02:11:57.920 --> 02:12:01.997
urgency demanded of this moment,

02:12:02.080 --> 02:12:06.157
ensuring that the most
powerful technological

02:12:06.240 --> 02:12:08.477
transition since the industrial

02:12:08.560 --> 02:12:12.237
revolution serves as an enduring engine

02:12:12.320 --> 02:12:15.997
for shared prosperity and a pillar of

02:12:16.080 --> 02:12:20.397
international peace and security rather

02:12:20.480 --> 02:12:21.997
than an opportunity.

02:12:22.080 --> 02:12:25.117
for further division, fraction of our

02:12:25.200 --> 02:12:29.197
world and insecurity.

02:12:29.280 --> 02:12:33.197
I thank you for your kind attention.

02:12:33.280 --> 02:12:38.077
>> I thank the permanent representative
of Liberia for his statement. There are no

02:12:38.160 --> 02:12:42.157
more names inscribed on the list
of speakers. I would like to thank the

02:12:42.240 --> 02:12:47.640
speakers at this Security Council
meeting. The meeting is adjourned.

02:15:36.880 --> 02:15:41.400
Hi. Hi.

02:15:54.160 --> 02:15:57.280
This is

02:16:06.240 --> 02:16:09.240
iPhone.

02:16:51.920 --> 02:16:55.720
I had to send that.

02:17:08.960 --> 02:17:12.237
The 10,227th meeting
of the security council is

02:17:12.320 --> 02:17:14.237
resumed.

02:17:14.320 --> 02:17:19.277
I now give the floor to his excellency,
Mr. Antonio Tajani, Deputy Prime

02:17:19.360 --> 02:17:25.120
Minister and Minister for Foreign Affairs
and International Cooperation of Italy.

02:17:26.160 --> 02:17:30.077
Five years after the beginning of the
Russian aggression against Ukraine, the

02:17:30.160 --> 02:17:32.397
position of Italian government led by

02:17:32.480 --> 02:17:35.117
Prime Minister George Maloney is clear.

02:17:35.200 --> 02:17:37.277
We stand with Ukraine and his brave

02:17:37.360 --> 02:17:41.037
people that are resisting an an

02:17:41.120 --> 02:17:46.237
injustice war. Our goal is a just
and lasting peace. We strongly condemn the

02:17:46.320 --> 02:17:50.077
Russian attacks against civilian
and energy infrastructures in Ukraine. The

02:17:50.160 --> 02:17:54.797
increasing debris threats by Russia are
putting European security in danger. We

02:17:54.880 --> 02:17:56.877
expect that our solidarity to Germany

02:17:56.960 --> 02:18:00.317
from the incidents at leaping airport

02:18:00.400 --> 02:18:03.357
Europe will be remain

02:18:03.440 --> 02:18:06.717
united. It's key to avoid an escalation

02:18:06.800 --> 02:18:08.957
and we want to be very clear. Italy is

02:18:09.040 --> 02:18:12.077
not at war with the Russian Federation.

02:18:12.160 --> 02:18:15.837
Mr. President, >> Mr.
President Moscow will always find

02:18:15.920 --> 02:18:19.517
Italy ready to support dialogue and
the pursuit of diplomatic solutions. But

02:18:19.600 --> 02:18:21.277
Russia as a permanent member of this

02:18:21.360 --> 02:18:24.397
council has a responsibility that cannot

02:18:24.480 --> 02:18:28.720
be ignored. The time has come for