OpenAI and Anthropic CEOs to Address UN Security Council on AI Safety Amid Wave of Hacking Incidents

Altman: catastrophe risk at '10% or 1%' is 'not remotely acceptable'

Altman built his remarks around two failure modes, framing the probability argument directly: 'It doesn't matter whether people put their risk of catastrophe at 10% or 1%, none of these levels are remotely acceptable.' He appeared live in New York at the Council session.

Altman named power concentration as a second AI failure mode alongside RSI

Alongside recursive self-improvement, Altman explicitly cited concentration of power 'in any company or country that believes only it can be trusted' as a second core risk — adding a geopolitical dimension to his technical risk framing at the Council.

Amodei called AI 'the most important global security issue facing the world today'

Amodei, appearing by video link, was more direct than Altman on governance, calling AI 'the most important global security issue facing the world today' and presenting a detailed playbook: simple treaties, global verification systems, and shared testing standards.

Discourse mapped as three poles: Anthropic, White House, and OpenAI in the middle

AI Daily Brief's framing of the Council session: Amodei/Anthropic at one governance pole (strong international controls), the White House at the other (reject global governance structures), and OpenAI charting a middle path — the clearest summary of where institutional positions landed.

Altman framed AI as a choice: 'new Renaissance of discovery' vs 'new Industrial Revolution of upheaval'

In his opening remarks at the Council, Altman presented AI's trajectory as a binary choice — a Renaissance that expands human creativity and agency, or a disruptive Industrial Revolution of upheaval and disarray. He emphasized AI should give people 'more agency in an ever more complex world,' not automate human agency away.

Altman cited two AI failure modes, not three, departing slightly from Bengio's framing

Altman said he 'largely agrees with Professor Bengio' but structured his risk framing around two failure modes rather than three steps. The first: AI moves so fast people can no longer follow or intervene. The official OpenAI transcript is the primary source for his full remarks.

Russian AI drone autonomously killed three Ukrainians in targeting decision last month

A Russian drone 'seemingly using AI' made an autonomous targeting decision that killed three Ukrainians last month, per European leaders at the Council. Zelenskyy cited this as proof that AI deciding battlefield outcomes 'is already a real possibility,' not a future concern.

UK FM Miliband committed to AI governance as G20 priority under UK presidency

UK Foreign Secretary Ed Miliband said when the UK takes the G20 presidency next year, it will put controlling AI 'at the heart of discussions,' seeking a single set of global AI safety standards.

White House adviser explicitly rejected new global AI governance structures

White House Science Adviser Michael Kratsios told the Security Council that concerns about AI control are 'not a reason to pause its further development or to constrain it with new global governance structure,' directly pushing back against the CEOs' calls.

Amodei proposed 3-step AI slowdown plan at the Security Council

Amodei's plan: (1) nations agree AI will not be used to build biological weapons; (2) develop methods to verify model capabilities; (3) establish common testing standards, including coordination between democratic and authoritarian governments.

Upcoming IPOs at OpenAI and Anthropic complicate public slowdown commitments

Neither company has disclosed an IPO timeline, but both are reportedly preparing for 'blockbuster IPOs,' raising questions about whether a development slowdown is compatible with investor expectations.

France convened the meeting; Barrot compared AI to the atomic age and HAL 9000

French FM Jean-Noël Barrot, who convened Wednesday's meeting, compared AI to 'the dawn of the atomic age' and the HAL 9000 computer, calling for international regulation just as nuclear weapons required it.

Zelenskyy warned AI may already decide what happens on the battlefield

Zelenskyy told the General Assembly: 'There is already a real possibility that AI, not only people, will begin to decide what happens on the battlefield. And we need peace before we reach that point.'