Public transcript

WATCH NOW: Anthropic CEO Dario Amodei Warns UN Security Council of AI Risks to Humanity | AI15

By: DWS News
Length: 5 min
Language: English
Text from: the video's captions
Transcribed: 13 October 2026

Read the full transcript Open on YouTube

Notes in English.

Summary

Anthropic CEO Dario Amodei briefs the UN Security Council on the rapid progress of AI, citing both its benefits, such as a Claude-led scientific discovery, and two major risks: misuse (including biological weapons) and loss of control. He argues AI safety is a global security issue beyond any one company, and proposes narrow international agreements, evaluation and verification systems, and common global testing standards plus an incident notification system.

Key points

Amodei says AI has advanced from barely writing code four years ago to writing most of the code at Anthropic and solving famous unsolved math problems.
He predicts one or two years, maybe less, to reach what he calls a country of geniuses in the data center.
Anthropic announced a Claude-led discovery of a new molecular machine that may constitute a new gene-editing mechanism with potential gene therapy applications.
He identifies two main risks: misuse of AI models, such as by bioterrorists, and loss of control as capabilities outpace developers' ability to manage them.
Anthropic tests every model for dangerous capabilities, publishes risk reports of hundreds of pages, and pioneered mechanistic interpretability and alignment research.
On September 12th he called for a collaborative effort to pace the frontier, including embedding external evaluators with employee-like access, similar to a food inspector.
He proposes narrow agreements such as a ban on using AI to make biological weapons, evaluation and verification systems, and common global testing standards with an incident notification system.
He calls AI the most important global security issue facing the world today, saying no leader, company, or nation can manage it alone.

Questions it answers

00:00Opening and AI progress

How quickly is AI advancing according to Dario Amodei?

He says AI went from barely writing code or solving high school math four years ago to writing most of Anthropic's code and solving famous unsolved math problems, and that one or two more years, maybe less, could reach a country of geniuses in the data center.

AI progress is extremely rapid and accelerating.
Amodei predicts a 'country of geniuses in the data center' within one or two years, maybe less.

00:45Benefits of AI

What benefits of AI does Amodei highlight?

He cites Anthropic's announcement of a Claude-led discovery of a new molecular machine, possibly a new gene-editing mechanism with gene therapy applications, and predicts profound near-term progress on diseases that have plagued humanity for millennia.

Claude led literature review, theorizing, and experimental design; humans ran and verified experiments.
He expects such applications to accelerate rapidly in the years ahead.

01:28Two categories of risk

What risks does Amodei see in AI?

He identifies misuse of AI models, such as by bioterrorists to create biological weapons, and loss of control, where model capabilities accelerate beyond developers' ability to control them. He says if managed poorly, AI could be a risk to humanity as a whole.

Misuse includes biological weapons creation.
Loss of control means capabilities outpacing developers' ability to manage them.
These are global issues concerning all of humanity.

02:11Anthropic's safety record

What does Anthropic do to manage AI risks?

Anthropic tests every model for dangerous capabilities in cybersecurity and biology, applies strong safeguards to its most capable models, publishes its research, pioneered mechanistic interpretability and alignment science, and releases risk reports of hundreds of pages. It commits to slowing down as much as necessary to ensure safety.

Anthropic claims the industry's best safety track record.
It will slow down as much as necessary to keep releases safe.

02:57Call to pace the frontier

What did Amodei propose on September 12th?

He called for a collaborative effort to pace the frontier — not stopping AI progress but ensuring safety keeps up with capability — with three parts: embedding external evaluators inside Anthropic with employee-like access like a food inspector, industry cooperation on standards and pace, and global government cooperation on international standards.

External evaluators would have employee-like access, similar to a food inspector.
Some companies have already agreed to adopt this measure.
International standards are the part that falls within the Security Council's remit.

04:39Proposals to the Council

What three ideas does Amodei offer the Security Council?

First, begin with narrow agreements all members can support, such as a ban on using AI to make biological weapons. Second, build evaluation and verification systems so states have visibility into frontier model capability and can verify commitments. Third, establish common global standards for testing loss-of-control and misuse risks, plus a notification system for AI incidents significant to global security.

Narrow agreements first, then broader verification systems.
A global incident notification system is proposed.
He calls this the most important global security issue facing the world today.

Notes

Context

Dario Amodei, CEO of Anthropic, briefs the UN Security Council at a meeting convened by France.
He frames AI risks as grave enough to belong before the Security Council, not just economic or ICT ministries.

Rapid AI Progress

Four years ago AI could barely write a line of code or solve a high school math problem; today it writes most of the code at Anthropic and solves world-famous unsolved math problems.
He estimates one or two years, maybe less, to reach a "country of geniuses in the data center."
Anthropic announced a Claude-led discovery of a new molecular machine, possibly a new gene-editing mechanism with gene therapy applications; Claude led literature review, theorizing, and experimental design while humans ran and verified experiments.

Two Main Risks

Misuse of AI models, for example by bioterrorists to create biological weapons.
Loss of control: model capabilities accelerating beyond developers' ability to control them.
If managed poorly, Amodei believes AI could be a risk to humanity as a whole.

Anthropic's Safety Practices

Tests every model for dangerous capabilities in areas like cybersecurity and biology.
Puts strong safeguards on its most capable models and publishes its work.
Pioneered mechanistic interpretability and alignment science; releases risk reports running to hundreds of pages.
Commits to slowing down as much as necessary so each successive AI technology released is safe.

Global Cooperation Proposals

On September 12th, Amodei called for a collaborative effort to "pace the frontier" — not stopping AI progress, but ensuring safety keeps up with capability.
Three parts: embedding external evaluators with employee-like access (like a food inspector), industry-wide cooperation on standards and pace, and global government cooperation on international standards.
To the Council he proposes: narrow agreements such as a ban on using AI to make biological weapons; evaluation and verification systems giving states visibility into frontier model capability; and common global testing standards for loss-of-control and misuse risks, plus notification of AI incidents significant to global security.

Conclusion

He calls this the most important global security issue facing the world today and urges members to set aside differences; no leader, company, or nation can manage it alone.

Text from the video's own captions. Summary and notes written by AI from the transcript, so check anything important against the recording.