Why the Heads of Top AI Companies Went to the UN
Sam Altman of OpenAI and Dario Amodei of Anthropic spoke to the United Nations Security Council on September 23, 2026, warning that artificial intelligence poses risks to humanity and asking governments to work together. Associated Press Reuters
The session was a Security Council meeting on artificial intelligence held during the 81st session of the UN General Assembly at UN headquarters. Associated Press Reports of the briefing named Altman and Amodei as the industry speakers addressing the Council. The Guardian
Altman and Amodei stressed global cooperation on standards for AI safety in their remarks. Bloomberg They pointed to the possibility that AI systems could gain the ability to improve on their own. Bloomberg
The appearance followed a paper from OpenAI. On September 21, 2026, OpenAI published a page titled "Building standards for the next phase of AI" outlining a path to shared global AI standards. OpenAI The outline calls for coordinated evaluation, reporting, and governance to improve AI safety. OpenAI
OpenAI listed 'Sam Altman's remarks at the United Nations Security Council' as a Global Affairs news item dated September 23, 2026. OpenAI
The broader context here is simple. Think of it like food safety checks. Evaluation is the test. Reporting is sharing the result with the regulator. Governance is the power to require fixes or pull a product. You need all three. Tests mean little if no one sees the results. Results mean little if no one can act.
Looking at what this means in practice, self-improvement makes timing matter. If an AI can change itself, a one-time test goes out of date fast. That puts more weight on steady monitoring and clear rules for when to retest. Rules help. They do not remove risk.
In my view, the location matters as much as the words. A Security Council briefing frames AI safety as a problem countries must solve together, not just a business issue or a race between nations. Shared rules make it harder to shop for weak oversight. They also lower the cost of working across borders. The hard part is speed. Agreement is slow. AI moves fast. Any workable plan must spell out what proof is needed, who gets to check the systems, and what happens if rules are broken. Without that, cooperation is an idea, not a system.
For people following this, the signal to watch is whether testing, reporting, and enforcement are agreed as one package. Scattered pilots will say little. Shared commitments across countries will say a lot.
