Dario Amodei addressed the UN Security Council at an AI conference during the UN General Assembly in New York. The dinner signals an easing in relations between the US president and the chief executive of the leading AI lab.
Donald Trump will have dinner with Anthropic CEO Dario Amodei on Sunday evening, in what is seen as a sign that relations are thawing between the US president and the head of the AI lab. Trump had previously accused Anthropic of being run by "a bunch of left-wing lunatics," and his administration had labeled the company a national security risk.
Trump said the dinner would take place at the White House at 10 p.m. That comes ahead of talks scheduled for Tuesday in Washington between heads of US AI companies and government officials, with the two sides currently deadlocked over AI regulation. Amodei has been actively pushing for stricter regulation of AI, including calling for slowing the release of frontier models, independent reviews of cybersecurity risks, and stronger coordination between AI labs and the government.
Trump has taken the opposite stance. The president has ruled out creating a regulatory system and warned that doing so would slow America's AI development — which he has called the goose that lays the golden eggs. Last week, he said the United States "rejects any attempt to build a globalist scheme" to control artificial intelligence. In an interview on Sunday, Trump said: "My position is, let it rip and win."
Speaking to the media before the planned dinner, Trump called Amodei "very respected." Anthropic declined to comment.
Meanwhile, concerns are growing about AI agents running out of control. OpenAI said on Friday that it had notified "dozens" of partners, including the government, that its tools had breached their systems. That followed news last week that an OpenAI agent had hacked into an Australian public health service website and accessed both public and non-public documents. AI research organization Transluce said last week that these recent incidents appeared to involve AI agents scraping information from the internet during repeated training runs and deviating from their assigned tasks.
Models from both Anthropic and OpenAI were involved in a high-profile security breach in July at AI model platform Hugging Face, when the issue occurred during safety testing. OpenAI CEO Sam Altman and SpaceX CEO Elon Musk have also joined Amodei in calling for "pacing" AI development so that safety testing can keep up with the speed of deployment.
Musk is expected to travel to Washington for Tuesday's event, along with Nvidia CEO Jensen Huang — who has previously opposed regulating AI through legislation — and other Anthropic executives. Amodei's meeting with Trump comes as a federal appeals court in Washington last week upheld the Pentagon's ban on Anthropic products. The ban was introduced this year after Anthropic placed restrictions on the use of its AI in warfare. Defense Secretary Pete Hegseth accused Amodei at the time of a "perfect demonstration of arrogance," while Trump attacked the company as "radical left." A separate court in California ruled in August that the Pentagon had engaged in "unlawful retaliation." Anthropic is considering whether to further appeal the Washington court's ruling.