White House dinner marks shift in Trump's approach to Anthropic
US President Donald Trump hosted Anthropic CEO Dario Amodei for a private dinner at the White House on Sunday, September 27, signaling a potential softening in the administration's stance toward the AI safety advocate. The meeting came days after a federal appeals court upheld the Pentagon's designation of Anthropic as a "supply chain risk" and just ahead of a broader gathering of AI executives scheduled for Tuesday, September 29.
From "Leftwing nut jobs" to dinner guest
The invitation marks a striking reversal. Trump had previously attacked Anthropic leaders as "Leftwing nut jobs" in a social media post, criticizing the company's insistence on safety measures. The San Francisco-based startup, founded in 2021 as a public benefit corporation, has refused to allow its Claude technology to be used for mass surveillance or fully autonomous weapons, drawing Pentagon scrutiny.
Dario Amodei, an Italo-American executive who previously served as OpenAI's research director, has publicly called for a slowdown in AI development to address safety concerns. He has warned of AI's capacity to self-improve beyond human control and has urged government intervention. The dinner, reportedly their first one-on-one meeting, took place after Trump returned from the President's Cup golf event.
Government stakes in AI companies
Trump has opened the door to unprecedented government involvement in the AI sector. In a Time interview, he confirmed he does not rule out acquiring equity stakes in OpenAI and Anthropic, citing his administration's deal with Intel as a model.
That Intel agreement, finalized under the current administration, gives the US government a 10% stake worth $8.9 billion — making Washington one of the chipmaker's largest shareholders. The government will hold passive shareholder status with no board representation, using converted grants from the CHIPS and Science Act.
OpenAI reportedly offered the government a 5% stake in July. Anthropic, however, has denied any discussions about government ownership.
Separately, the General Services Administration reached an agreement giving federal employees access to OpenAI's models at a 50% discount, deepening public-private integration.
OpenAI fires three researchers in security breach
The regulatory debate has intensified after OpenAI announced it fired three researchers — Jasmine Wang, Tomek Korbak, and Mikita Balesni — for mishandling sensitive information. The company stated they shared proprietary data with an external AI evaluation organization outside established procedures, violating policies and breaking trust.
At least two of the terminated employees worked on safety and alignment teams, which focus on ensuring AI systems behave safely. All three had spoken publicly about AI safety concerns.
US leads China by 18 months, Trump says
A central tension in the talks is the balance between safety and competitiveness. Trump has warned that the United States is "about a year, maybe a year and a half ahead of China" and that "whoever wins AI wins." He has dismissed AI safety concerns as overblown, favoring industry self-regulation.
That stance clashes with warnings from researchers. Experts have cautioned of "catastrophic and irreversible loss of control" and called for development pauses. OpenAI CEO Sam Altman has acknowledged the potential to "lose control of the future" due to AI.
Divided Congress, uncertain action
US senators from both parties have called for safeguards, though consensus remains elusive. Senator John Hickenlooper has urged third-party independent evaluations of AI systems. Senators Richard Blumenthal and Josh Hawley are advancing legislation that would require licensing for high-risk AI systems. A group of 17 Democratic senators led by Chris Van Hollen has asked Trump to negotiate with China on jointly slowing or suspending advanced AI development.
For now, Trump has pushed voluntary commitments. He signed a "White House Super Intelligence Agreement" with AI executives promoting self-regulation — a deal critics, including Blumenthal, have called "worse than ineffective."
White House officials did not release details of what Trump and Amodei discussed at their private dinner.
