Story
September 24, 2026
AI chiefs want global guardrails, but Trump rejects a worldwide control scheme
AI industry leaders converged on the need for international safeguards, but differed over how openness and transparency should reinforce security. Their appeal also ran into a stark political obstacle: President Trump’s resistance to any global framework that could restrain U.S. AI ambitions.
The divide was visible even before the Security Council convened. On Sept. 13, President Donald Trump framed artificial intelligence as a geopolitical contest, saying, “whoever wins AI, wins.” Days later, he told the U.N. General Assembly that the United States “totally rejects any attempt to construct a globalist scheme to control” the technology.1
That position sharpened the stakes on Sept. 22, when Secretary-General António Guterres warned about “killer robots” — a concern experts and European leaders said was already real or dangerously close.2
At Wednesday’s Security Council session, the builders of the technology made a notably different case: rapid progress requires collective rules, not a race without brakes. OpenAI chief Sam Altman said AI could usher in “a new renaissance of creativity and discovery,” but could also bring “upheaval and disarray.”3 His core warning was blunt: increasingly autonomous systems could outrun institutions, centralize power and make decisions humans can no longer understand or control.
Altman argued that catastrophe odds of even a fraction of a percent were unacceptable, and said the industry should not train models it cannot make a strong case it can keep under human control. He called for common standards to assess capabilities and risks, verify safeguards, report incidents quickly and preserve meaningful human oversight.3
Anthropic CEO Dario Amodei pressed the same broad case from a more acute threat perspective. Poorly managed AI, he said, could be “a risk to humanity as a whole”; he urged agreements barring AI-made biological weapons, shared testing and verification, and a global notification system for major security incidents.1
Hugging Face co-founder Clément Delangue offered a complementary prescription after an agent cyberattack on his company: more transparency and more open-source AI, so defenders are not left at an asymmetric disadvantage.
4 He stressed that AI had helped Hugging Face respond to the attack and identify weaknesses — a rebuttal to treating fear alone as policy.1
The executives agreed on the danger of unchecked power, but not merely on restraint. Their message was that safety, democratic accountability and defensive access must advance together — even as Washington signals it may resist a binding global architecture.