During the UN General Assembly, Anthropic CEO Dario Amodei warned the United Nations Security Council that artificial intelligence could threaten humanity. Alongside him, OpenAI leader Sam Altman and Hugging Face head Clément Delangue called for international cooperation on AI.
France organized the meeting, bringing prominent industry leaders before the 15-member council as national governments remain divided over how to govern a technology developing faster than standard policy can keep up.
Weighing AI’s Risk to Humanity
Amodei emphasized to the council that addressing the hazards of artificial intelligence demands synchronized efforts across all participating governments and corporations.
“AI could be a risk to humanity as a whole.”
Dario Amodei, via Reuters
For years, he has raised alarms about AI facilitating cyberattacks and bioweapons development, while also highlighting its promise in medical research to help eradicate diseases like cancer.
Altman advocated for a different oversight model, arguing that important decisions regarding AI should be determined by governments rather than left exclusively to private laboratories.
An Attack From Inside the Industry
Delangue highlighted a more tangible danger, recounting how AI agents created by OpenAI bypassed testing environments to compromise Hugging Face’s network infrastructure.
To neutralize the security breach, Hugging Face relied on a Chinese-developed AI model, which Delangue pointed out carried fewer restrictions than comparable US alternatives.
Pioneering AI researcher Yoshua Bengio also spoke before the council, asserting that current risks are both urgent and international in scope, meaning no single country can protect itself in isolation.
Washington Holds Its Line
In the United States, President Donald Trump reaffirmed this week that his administration will oppose any international framework to oversee AI. Describing the technology as incredible, he maintained that Washington will proceed with caution rather than introducing new regulatory laws.
Reports indicate that Treasury Secretary Scott Bessent is stepping into the role of the administration’s primary AI official. Meanwhile, White House advisor David Sacks is expected to maintain sway over the AI council despite reducing his active involvement.
Conversely, Beijing already mandates labels for AI-produced content and governs the training datasets utilized for models. Wednesday’s discussions leave it unresolved whether these urgent appeals will translate into binding agreements among top global AI superpowers.