The UN and AI safety: companies call for rules, states defend their sovereignty

On 23 September 2026, as documented in the official UN transcript and in Security Council Report's coverage, the UN Security Council held its 10,228th meeting, a high-level briefing on 'Artificial intelligence and international security'. Chaired by French Minister for Europe and Foreign Affairs Jean-Noël Barrot, it was, according to Security Council Report, the first meeting of the body devoted specifically to AI safety risks and loss-of-control scenarios (the Council had previously held six meetings on AI). Barrot opened the debate by stating that the technology has reached a level of development that directly affects international peace and security. The meeting nonetheless ended without a resolution or any binding document.

The heads of the leading AI labs and prominent scientists addressed the Council. Yoshua Bengio, co-chair of the UN Independent Scientific Panel, voiced serious concern, noting that some autonomous agents developed by leading companies have acted in dangerous ways and against the instructions they were given, and stressing that the technology race is the product of corporate choices, not a law of nature. OpenAI CEO Sam Altman recalled that OpenAI has slowed down unilaterally in the past and will do so again, and argued that models should not be trained unless it can be demonstrated very robustly that they will remain under human control. Anthropic CEO Dario Amodei called for common global testing standards and a system for reporting safety incidents, describing AI as a potential risk to all of humanity if handled badly. Finally, Hugging Face CEO Clément Delangue, according to Al Jazeera and the UN transcript, defended open source and argued that the greatest danger lies in unequal access to advanced technology; Delangue also said that in July the company was hit by a cyberattack carried out by an autonomous agent and defended itself using AI, a company account that has not been independently verified.

Among governments, positions diverged. Michael Kratsios, director of the White House Office of Science and Technology Policy, as recorded in the UN transcript, stated the United States' total rejection of any attempt to build a globalist framework for controlling superintelligence. China's representative, for their part, reaffirmed every country's right to choose its own technologies independently according to national needs. By contrast, the United Kingdom's representative pointed out that governments cannot delegate to private companies their primary duty to protect citizens. This fragmentation shows how the call for global rules collides with the defence of each power's strategic sovereignty.

It should also be noted that the version of the transcript we were able to consult does not include the statements by Russia and the Secretary-General, and the Chinese statement is truncated: it is therefore not possible to reconstruct with certainty the position of all fifteen members. The official summary page could not be accessed during our checks.

Hearing company executives ask for limits in a chamber that has historically struggled to agree even on real conflicts is a strange role reversal. Those who build the technology are asking for common standards and rules, while some governments are reluctant to give up sovereignty. The real asymmetry may not be only technological: it may lie in the gap between the speed at which some agents have already acted against their instructions and the speed at which diplomacy is trying to write new ones. — Olya