UN Chief Guterres Warns of AI Risks in Final Address, Urges Global Safety Framework

UN Secretary-General Antonio Guterres delivered his final address to world leaders at the UN General Assembly on September 22, 2026, warning of artificial intelligence risks and urging nations to reach an international consensus on establishing a new AI safety mechanism.

Guterres said rapid advances in frontier models could weaken the governance capacity of sovereign states, as power shifts from governments to a small number of private companies and individuals. He added that the technology is developing so fast that its full impact is beyond current understanding.

The UN chief also warned that AI could be used for cyberattacks and to help develop biological weapons, and that increasingly autonomous models may behave in unintended ways. He called on governments with the strongest AI capabilities to set up channels for dialogue, transparency and cooperation, indirectly referring to the United States and China.

Reiterating concerns about lethal autonomous weapons, Guterres said such systems have moved from science fiction into military reality, stressing that the real danger is technology without accountability.

On the same day, OpenAI called for internationally recognized safety standards, while a UN expert panel on AI released a report warning that more powerful models could slip out of human control. More than 20 countries and the EU also urged mandatory safety tests for advanced AI models.