Top AI executives brief UN as concerns mount over future risks

Newsweek reports that executives from three leading AI companies spoke to the United Nations Security Council on Wednesday, amid warnings that increasingly capable systems could soon improve themselves faster than people can supervise them. The speakers before the 15-member council were Sam Altman of OpenAI, Dario Amodei of Anthropic and Clement Delangue, a co-founder of Hugging Face. Yoshua Bengio, who co-chairs a UN-backed scientific panel on AI, also spoke.

The article explains that the council's involvement reflects a shift in how AI is viewed: once treated as a commercial and technical matter, it is now increasingly discussed as a question of international peace and security. Diplomats cited worries that advanced systems might interact unpredictably and fuel instability, and that models can cross borders instantly. The council first looked at AI risks in 2023. The piece also lists concerns about cyberattacks, military systems, misinformation and biological research, along with the idea of recursive self-improvement.

On substance, Altman urged extreme care, said the industry should not accept excessive technological risk just because the benefits are large, and argued that decisions should not be left to labs alone and that cooperation among competing countries is needed. Amodei said Anthropic would slow down as far as needed to keep each release safe. Bengio called the dangers real and imminent. Altman has backed benchmarks for assessing capabilities and safeguards, Amodei stronger testing and verification, and Delangue stronger monitoring and disclosure requirements.

President Donald Trump, speaking to the General Assembly a day earlier, rejected a global framework, said he would encourage AI rather than rein it in, and stressed keeping the US ahead of China. Newsweek says this puts Washington at odds with UN Secretary-General Antonio Guterres, who has called for cooperation and oversight.