FULL STORY

AI Chiefs Address UN Security Council for First Time

Altman, Amodei and Bengio addressed the UN Security Council's first AI session, where Altman pitched US-led global AI standards and cast himself as a centrist, drawing pushback from safety researchers.

2026-09-21 ~ 2026-09-26 · 6 episodes · 40 posts

Episode 1 · Altman and Amodei to Address UN Security Council on AI Safety (2026-09-21, 6 posts)

OpenAI CEO Sam Altman and Anthropic CEO Dario Amodei will address a special UN Security Council session on AI, joined by Turing Award winner Yoshua Bengio and Hugging Face CEO Clément Delangue. Altman plans to advocate shared global AI safety standards while avoiding measures that slow innovation, marking a rare appearance of top AI executives before the Security Council.

Confirmed

  • Altman will call for global AI safety standards at the Council; OpenAI says it does not back a specific international regulator, favoring baseline safety standards plus democratic oversight.
  • Amodei will participate via video link.
  • Other speakers include Yoshua Bengio and Hugging Face CEO Clément Delangue.
  • Altman intends to position himself as a moderate on whether AI development should slow down.
  • The session, covering AI risks and implications for international security, is scheduled for Wednesday (per @GaryMarcus citing Kalshi).

Unconfirmed

  • An exclusive from @terryyuezhuo claims DeepSeek will brief the Council on AI risks and international security and that Moonshot has also been invited; this remains single-sourced and unverified.

Why it matters

  • A rare AI-focused Security Council session gathering major AI executives and academics could set the tone for global AI governance debates.
  • OpenAI's explicit opposition to a dedicated international regulator reveals the tug-of-war between industry and regulators over regulatory design.
  • AI critic and NYU professor Gary Marcus mocked Altman as speaking with "no conflicts of interest at all," underscoring skepticism about the independence of corporate leaders in policy discussions.

Episode 2 · OpenAI Calls for US-Led Global AI Safety Standards (2026-09-22, 9 posts)

On September 22, OpenAI published "Building standards for the next phase of AI," urging the US to lead a global AI safety and assurance standards framework with international coordination. CEO Sam Altman added that the public outside AI labs should have real voice in the technology's direction and a clear way to verify safety commitments, and that standards should help prevent power concentration, including keeping new and open-model companies competitive. The move is seen as a major step in OpenAI's policy positioning, with top labs actively pulling governments into the standard-setting process.

Confirmed

  • The core proposal calls for US leadership in setting global AI safety and assurance standards, covering capability evaluations, risk assessment, human oversight, and a unified incident-reporting mechanism
  • Per @JacquesThibs, three key points: international standards for recursively self-improving (RSI) systems; linking national AI safety institutes into a unified network; systematically tracking RSI-related progress and each lab's internal autonomous research
  • Standards should track RSI-related capability gains and the share of autonomous research inside labs, with thresholds; once autonomous research exceeds a threshold, immediate human review is required
  • Per @ShakeelHashim relaying reporter Zeff Max, OpenAI is asking the US government to lead the global standards effort
  • Per @emmanuelvivier, SoftBank issued over $11 billion in bonds to fund its investment in OpenAI

Why it matters

  • Altman's framing of "outside voice" and "preventing power concentration" directly responds to criticism of OpenAI's own influence and writes open-model and startup companies into the competitive framework
  • The proposal puts RSI and autonomous-research ratios on the international tracking agenda, touching the most sensitive capability boundaries in current AI safety debates
  • @TansuYegen notes the timing is striking: the company running fastest in the race is also the loudest voice calling for rules — not necessarily insincere, but whoever writes the standards may shape the industry as deeply as whoever builds the strongest model

Episode 3 · Altman, Amodei and Bengio Address UN Security Council on AI Governance for the First Time (2026-09-22, 19 posts)

On September 24, the UN Security Council (its 10228th meeting, on AI and international security) held its first session on artificial intelligence. OpenAI CEO Sam Altman and Anthropic CEO Dario Amodei testified, alongside Turing Award winner Yoshua Bengio, co-chair of the UN's independent scientific AI panel. Frontier lab leaders made a rare joint appearance at the highest level of international governance to call for coordinated global AI rules, though critics pushed back.

Confirmed

  • Altman warned that as systems become self-improving and AI-building grows automated, progress could accelerate dramatically and humanity could "lose the future to AI" — moving too fast for people to follow or intervene when needed. He said no individual, company or country should be allowed to monopolize the most powerful AI, calling for "extreme care" at this moment.
  • Altman cautioned that a company believing only it can be trusted with the most powerful AI would use that belief to justify nearly anything, and that a 0.1% chance of catastrophe should be no more acceptable than 10%. Poster victorexplore noted this effectively describes the narrative every AI vendor sells.
  • Altman disclosed that OpenAI has unilaterally slowed its AI development in the past and would do so again.
  • He also sought middle ground, rejecting both blind optimism and doomerism, comparing some peers' rhetoric to "a bad sci-fi movie."
  • Per CNN and the New York Times, Altman and Amodei urged the Council to adopt international AI standards and regulatory frameworks, arguing AI risks are not unfounded and that without global governance a crisis looms.
  • Amodei proposed that Anthropic has committed to embedding external evaluators — like "food inspectors" — inside the company with near-employee system access, and suggested other AI firms worldwide do the same. He touted Anthropic's safety record; commenter teortaxesTex quipped he was marketing his own practices on the Council stage.
  • Bengio warned of an "unprecedented threat" from uncontrollable frontier AI agents, likening the moment to historical inflection points where humanity successfully managed catastrophic risks.

Unconfirmed

  • Altman's claim of having deliberately slowed development rests solely on his own account, with no project or timeline specified.

Why it matters

  • This was the first time the two frontier lab chiefs appeared before the Security Council on the same day, marking a shift toward top-level international regulatory engagement; Amodei's resident-evaluator proposal offers a concrete regulatory template.
  • Critics responded in parallel: Gary Marcus mocked having Altman brief on safety as "an arsonist writing the fire code" and "Al Capone teaching the FBI," underscoring sharp industry divides over who gets to define the AI safety agenda.

Episode 4 · Altman's AI Governance Proposal Draws Sharp Criticism from Safety Researchers (2026-09-23, 2 posts)

Sam Altman published an AI governance proposal calling for real outside oversight of labs, clear safety standards, and protections against power concentration. Safety researchers pushed back point by point, and critics questioned the timing of the post.

Episode 5 · Altman Tells UN Security Council Global AI Rules Can't Be Made in San Francisco Alone (2026-09-24, 2 posts)

Sam Altman pitched global AI standards at the UN Security Council, positioning himself as a moderate and saying the most important AI decisions can't be left to San Francisco labs alone. E/acc figures mocked the move.

Episode 6 · Altman and Amodei Address UN Security Council on AI Risks (2026-09-25, 2 posts)

Sam Altman and Dario Amodei told the UN Security Council that AI is now the world's top security issue, with Amodei calling for a ban on AI-engineered bioweapons and international verification mechanisms.