Anthropic CEO shifts tone after White House dinner: AI poses risks, but the US must 'win safely'

Beating AI News Flash: Two weeks after publicly calling for the AI industry to "hit the brakes," Anthropic CEO Dario Amodei's tone has clearly shifted after two meals with Trump. Following a White House lunch, he still acknowledged that AI poses "very real risks," but he no longer mentioned slowing down model development. Instead, he quoted Trump, saying, "Whoever wins AI, wins."

Amodei subsequently stated that how to address these risks is "still under discussion," but everyone needs to work together to ensure the U.S. can both win and "win safely." He then deliberately added: "If we work with the president and everyone here, we can win safely."

This wording stands in stark contrast to two weeks ago. At that time, Amodei directly wrote, "We must slow down the pace of improving AI model capabilities," arguing that continuing at the current speed could lead to more severe runaway risks in the coming months. Trump later publicly opposed slowing down, saying doing so would only let China catch up to the U.S., and dismissed some AI disaster warnings as a "hoax."

The two, who previously had tense relations, sat at the same table twice this week. Amodei first dined alone with Trump on Sunday evening, and two days later attended the White House AI lunch, signing a voluntary safety agreement alongside Google, Meta, OpenAI, xAI, and Nvidia. The risks remain, slowing down is no longer mentioned, and the new keyword has become "win safely."