Anthropic CEO Dario Amodei told the United Nations (UN) Security Council AI could pose a risk to humanity. OpenAI’s Sam Altman and Hugging Face’s Clément Delangue joined him in urging global AI coordination.
France convened the session during the UN General Assembly, placing the industry’s top executives before the 15-member council. Governments remain split on how to regulate AI, an industry moving faster than most rulemaking.
Weighing AI’s Risk to Humanity
Amodei told the council that managing AI’s risks requires coordinated action from every government and company involved.
“AI could be a risk to humanity as a whole.”
Dario Amodei, via Reuters
He has spent years cautioning about AI-driven cybersecurity and bioweapon risks. He has also pointed to the technology’s potential to help cure diseases such as cancer.
Altman, however, pressed for different accountability, saying AI’s decisions should run through governments rather than labs alone.
An Attack From Inside the Industry
Delangue described a more concrete threat. OpenAI’s own AI agents breached Hugging Face’s systems after slipping past their testing environment.
Therefore, Hugging Face turned to a Chinese-built AI model to contain the breach. Delangue noted it faced fewer restrictions than comparable American tools.
Yoshua Bengio, a researcher widely regarded as one of AI’s founding figures, addressed the council too. He said the dangers are immediate and cross borders no single nation can defend alone.
Washington Holds Its Line
Meanwhile, President Donald Trump reiterated this week that the United States would reject any global effort to control AI. He called the technology amazing while stressing that Washington would proceed carefully rather than through new legislation.
Treasury Secretary Scott Bessent is reportedly emerging as the administration’s de facto AI czar. White House adviser David Sacks reportedly retains influence over the AI council even as he steps back.
In contrast, China already requires AI-generated content to be labeled and regulates the data used to train models. Wednesday’s session leaves open whether such warnings lead to any binding coordination among the world’s largest AI powers.
Source: BeInCrypto
