A surprising shift in relations between the Trump White House and leading artificial intelligence developer Anthropic has unfolded, with senior administration officials holding a “productive and constructive” meeting with Anthropic CEO Dario Amodei — even as the AI firm continues its legal battle against the U.S. Department of Defense.
The high-stakes sit-down, which took place on Friday, brought Amodei face-to-face with Treasury Secretary Scott Bessent and White House Chief of Staff Susie Wiles, according to reporting from Axios. The gathering comes just one week after Anthropic launched the public preview of its latest cutting-edge AI system, Claude Mythos, a tool the company says exceeds human performance in a range of specialized hacking and cybersecurity tasks. To date, access to Claude Mythos has been restricted to only a small group of vetted organizations, but independent cybersecurity researchers have already confirmed the system’s unusual capabilities, describing it as “strikingly capable at computer security tasks.” Per Anthropic’s own documentation, Claude Mythos can independently detect hidden vulnerabilities in decades-old legacy code and autonomously develop working methods to exploit those flaws.
This new overture from the White House marks a dramatic reversal from just two months ago, when the Trump administration publicly derided Anthropic as a “radical left, woke company.” Just months earlier in March, the Department of Defense labeled Anthropic a “national supply chain risk,” marking the first time a major U.S. technology company had received this public designation, which bars the firm from new federal government contracts. In response, Anthropic launched a lawsuit against the Defense Department and multiple other federal agencies, challenging the designation in court.
Anthropic has been a key provider of AI tools for high-level U.S. government and military work since 2024, and the company argues the supply chain risk label is nothing more than retaliation from Defense Secretary Pete Hegseth. According to Anthropic’s legal filing, Hegseth retaliated after Amodei rejected the Pentagon’s demand for unfettered access to the company’s full AI model stack, over Amodei’s deep concerns that the technology could be misused for mass domestic surveillance and the development of fully autonomous lethal weapons. While a California federal court ruled largely in Anthropic’s favor in early proceedings, a federal appeals court recently rejected the firm’s request for an emergency temporary stay to block the supply chain risk designation. Even so, Anthropic’s tools remain in active use across all federal agencies that had already deployed the technology prior to the designation.
Prior to Friday’s meeting, the White House had not issued any positive public remarks about Anthropic, with President Donald Trump himself taking to social media earlier this year to call for an outright ban on the company’s technology across the federal government. In that post, Trump claimed Anthropic was led by “left wing nut jobs” attempting to “strong arm” the U.S. defense establishment, writing, “We don’t need it, we don’t want it, and will not do business with them again!” When reporters questioned Trump about Amodei’s White House visit during a public appearance in Phoenix, Arizona on Friday, the president told reporters he had “no idea” the meeting was taking place.
In an official statement after the meeting, the White House confirmed the two sides discussed pathways for potential collaboration, as well as shared frameworks and safety protocols to manage risks associated with scaling advanced AI systems like Claude Mythos. “We discussed opportunities for collaboration, as well as shared approaches and protocols to address the challenges associated with scaling this technology,” the White House said, adding that the meeting “explored the balance between advancing innovation and ensuring safety.” Industry analysts say the meeting makes clear that despite the Trump administration’s public hostility toward Anthropic, the company’s advanced AI capabilities — particularly in the cybersecurity space — have become too strategically critical for the U.S. government to cut ties entirely. A representative for Anthropic declined to provide any additional comment on the details of Friday’s meeting when contacted by reporters.