The White House and the Pentagon are on opposite sides of a growing fight over whether the federal government should use Anthropic's artificial intelligence systems. It's a dispute that pits national security concerns against the desire to harness cutting-edge AI tools — and President Trump just threw a wrench in the Pentagon's hardline stance.

Pentagon Moves to Restrict Anthropic

The Pentagon has blacklisted Anthropic from military work, citing concerns that its AI could be misused for surveillance or autonomous weapons. It ordered agencies to remove Anthropic's products within six months.

Defense Secretary Pete Hegseth didn't mince words. During a Senate hearing last week, he called CEO Dario Amodei an "ideological lunatic" and compared the situation to "Boeing giving us airplanes and telling us who we can shoot at." Pentagon officials have labeled Anthropic a "supply chain risk" — a designation usually reserved for foreign adversaries — and ordered its removal from defense systems, with limited exceptions.

But President Donald Trump has signaled a more flexible approach. In an interview with CNBC, he said a deal with Anthropic is "possible" and praised the company's technical talent. "They're very smart and I think they can be of great use," he said.

Jessica Tillipman, a government procurement law expert, summed up the confusion: "They've now adopted completely inconsistent positions across the government." She told The Hill, "One agency continues to dig its heels… and the rest is trying to actually work in reality."

US Split Over Anthropic AI Risk and Access

The rift runs deeper than just the Pentagon. Earlier, U.S. officials remained divided over Anthropic's AI systems, with the Pentagon labeling its Claude models a supply chain risk and considering phasing them out across agencies.

Pentagon tech chief Emil Michael acknowledged that Anthropic's Mythos model could help identify cybersecurity vulnerabilities but described it as a "separate national security moment" requiring broader government safeguards.

Meanwhile, White House officials were drafting an AI policy for national security agencies that promotes the use of multiple AI providers and explores ways to allow controlled access to Mythos despite the Pentagon's concerns. The dispute comes as Amodei has been engaging with White House officials to resolve tensions following the Pentagon ban and a lawsuit over the risk designation.

Despite the conflict, federal agencies are still considering limited access to Mythos under proposed safeguards — highlighting just how uncertain U.S. AI policy remains.