Public transcript

LIVE: Governing in the Age of AI | fireside chat with former UK PM Rishi Sunak, Meta’s Alexandr Wang

By: ANI News
Length: 1 h 14 min
Language: English
Text from: the video's captions
Transcribed: 12 October 2026

Read the full transcript Open on YouTube

Notes in English.

Summary

In this fireside chat, former UK Prime Minister Rishi Sunak speaks with Meta's Alexandr Wang about governing in the age of AI, covering the technology's accelerating development, public trust, government deployment, and national AI strategy. Wang describes 2026 as an inflection point driven by recursive self-improvement and working agents, and argues trust will determine diffusion more than the technology itself. Sunak closes by urging political leaders to treat AI as an immediate, centralized responsibility of government rather than tomorrow's issue.

Key points

Wang traces three eras of AI: pre-training (2018–2024), reinforcement learning and reasoning (from late 2024), and recursive self-improvement beginning at the end of 2025.
He predicts 2026 will be an inflection point with large-scale agent deployments across many areas of the economy.
Sunak notes large AI companies will spend 20 times more on the technology this year than the US spent on the Manhattan Project.
Wang says trust, more than the technology itself, will govern how successfully AI diffuses, citing WhatsApp's privacy commitment as a model.
Sunak highlights the UK's AI Security Institute, created after Bletchley Park, to do pre-deployment testing of frontier models in partnership with labs.
Wang's playbook for governments includes data sovereignty (especially healthcare and national security data), compute access, innovation ecosystems, and public sector use cases.
Wang advises organizations to become AI-first with top-down direction plus bottom-up success stories, noting some Meta engineers are 10 to 100 times more productive with AI.
Sunak argues AI must move from a specialized topic to a centralized responsibility of government with urgent action.

Questions it answers

00:00Why the AI summits began

Why did Rishi Sunak create a global AI summit?

He found there was no international gathering on the most transformative technology of the time, and believed industry and innovators needed to be in the room with governments, unlike the G7 and G20 format. This led to Bletchley Park and a series that continued in Paris, South Korea and India.

Large AI companies will spend 20 times more this year than the US spent on the Manhattan Project.
The summit brings industry, innovators and government together.

01:31Eras of AI development

How has AI development evolved in recent years?

Wang identifies pre-training (2018–2024), reinforcement learning and reasoning from late 2024, and recursive self-improvement from the end of 2025, where models help build the next AIs. He expects 2026 to be an inflection point with large-scale agent deployments.

Researcher productivity is growing dramatically and expected to keep accelerating.
Agents started actually working in mid-to-late 2025, beginning with coding agents.

06:22Trust in AI

How can trust in AI be built, especially in Western countries?

Wang says trust may matter more than the technology for diffusion, and points to WhatsApp's privacy commitment as the model. Sunak highlights the UK's AI Security Institute doing pre-deployment testing with frontier labs as a global public good.

India shows optimism toward AI while Western countries show anxiety.
Proper testing and monitoring systems for national security and cyber risks are of paramount importance.

12:51Public sector deployment

Where are the clearest AI use cases in government?

Wang sees governments as service providers with huge opportunity to deliver services faster, noting some Indian states already deliver most government services through WhatsApp. Sunak says the trust debate goes from abstract to real when benefits are demonstrated.

AI can move governments away from large-scale bureaucratic services toward faster delivery.
Healthcare and security are key areas.

16:20Diffusion and India's strategy

What can governments learn from India's approach to AI?

Citing Professor Jeffrey Ding's book, Sunak argues you don't need to invent a technology to benefit most from it. India focused on mass adoption using digital public infrastructure — ID, UPI, payments — plus a deep talent pool and supportive population.

India has more consumer AI startups than the United States.
Deployment and adoption strategy can establish technology leadership.

17:50The government playbook

What building blocks must governments get right for AI?

Wang's playbook: data sovereignty (especially healthcare and national security data), compute access via alliances or cloud partnerships, a supportive innovation ecosystem, and public sector use cases. Not every country needs its own large-scale AI data centers.

Data is the new oil; governments should map their data reserves.
Compute access can come through partnerships rather than domestic build-out.

22:12Driving execution

How should CEOs and governments drive AI adoption?

Wang recommends a mix of top-down clarity that the organization is AI-first, where success or failure rests on embracing AI, plus bottom-up success stories. At Meta, some engineers are 10 to 100 times more productive with AI.

Sunak argues AI must be driven personally by the person at the top.
The transition should be framed as mandatory.

25:46Future vision

What is Alexandr Wang most excited about going forward?

He expects development and diffusion to keep accelerating, and is excited about personal superintelligence — agents helping each person live their maximal life — and interpersonal superintelligence, where AI helps humans organize better to solve great challenges.

Diffusion may speed up two to three times per year.
Sunak closes that AI must move from a specialized topic to a centralized responsibility of government.

Notes

Context and Origins

Rishi Sunak explains why he convened the AI summit at Bletchley Park: there was no regular gathering of leaders on the most transformative technology of the time, and industry needed to be in the room alongside government.
He notes large AI companies will spend 20 times more on developing the technology this year than the US spent on the Manhattan Project.
The summit series has since gone to Paris, South Korea, and now India, a country with huge digital ambitions and capacity.

The Technology Journey

Alexandr Wang describes three eras: pre-training (2018–2024) with a predictable exponential scaling curve; reinforcement learning from late 2024, when models began to reason; and recursive self-improvement beginning at the end of 2025, where models accelerate the production of the next AIs.
He calls 2026 an inflection point and predicts large-scale agent deployments across many areas of the economy, with the "GDP of AI" growing exponentially.
Individual researcher productivity has grown dramatically and is expected to keep accelerating.

Trust

Sunak observes optimism and trust toward AI in India versus persistent anxiety in Western countries, framing the confidence gap as both a policy and technical task.
Wang says trust may matter more than the technology itself for diffusion, citing WhatsApp's privacy commitment as the model for AI products.
Sunak describes the UK's AI Security Institute, established after Bletchley, doing pre-deployment testing with frontier labs as a global public good.

Government Playbook

Wang urges governments to treat data as the "new oil," building sovereignty over data assets, especially in healthcare and national security.
Other layers: compute/infrastructure access via alliances or cloud partnerships, innovation ecosystems (India has more consumer AI startups than the US), and government services.
Sunak references Professor Jeffrey Ding's book "Technology and Great Powers": you don't need to invent a general-purpose technology to benefit most from it — India's deployment-focused strategy with UPI, ID, and digital public infrastructure proves the point.

Execution and Vision

Wang advises becoming an AI-first organization: top-down mandatory direction plus bottom-up success stories, with some Meta engineers 10 to 100 times more productive using AI.
His long-term vision is personal superintelligence — an agent helping each person live their maximal life — and interpersonal superintelligence, where AI helps humans organize to solve great challenges.
Sunak closes: AI is an "action this day" issue that must move from a specialized topic to a centralized responsibility of government.

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