Trump Picked A Fight With Anthropic. Now The Administration Is Backing Off.

The Trump administration is showing every sign of backing down from its noisy public feud with Anthropic.

Just two months after President Donald Trump declared a government-wide blacklist of the tech startup’s AI software, cybersecurity worries tied to the company’s powerful new Mythos model have both sides searching for a ceasefire, according to six tech lobbyists and policy experts, including two with ties to the White House.

The administration has also taken public steps to loosen restrictions on federal agencies’ use of Anthropic software — moves followed by a cordial White House meeting on Friday with Anthropic CEO Dario Amodei, and by Trump’s comments this week praising the company’s executives as “very smart people.”

Most recently, the Justice Department asked a judge on Wednesday to pause the government’s appeal of a ruling that had blocked much of the administration’s effort to punish Anthropic.

The Defense Department still hasn’t dropped its most aggressive attack on the company — Secretary Pete Hegseth’s March declaration that Anthropic’s ethics policies pose a risk to the national security supply chain. And the sides are still at odds in two federal lawsuits over the administration’s penalties.

But increasingly, the balance of power inside the White House has shifted toward a faction that wants to make peace, six people closely following the dispute told POLITICO. Three of the people were granted anonymity to speak candidly about the still-fraught relationship.

One major motivation for ceasing hostilities is to enable government agencies to use Anthropic’s Mythos AI model so they can find and plug key cyber vulnerabilities — before an adversary finds them first.

“The administration is definitely still fighting [Anthropic in court]. I wouldn't characterize it as being over in that regard,” said Dean Ball, a former White House adviser who served as lead author of the administration’s AI Action Plan. “But I think that clearly there are some people who see the whole thing as a distraction they don't need.”

Ball suggested that’s particularly true of those in the administration “who were only spectators” when the Defense Department began sparring with Anthropic in February over the company’s attempt to impose ethical limits on the military’s use of its software. Neither Hegseth nor Defense Undersecretary Emil Michael, who has led much of the department’s criticism of Anthropic, were present at Friday’s meeting. Instead, White House chief of staff Susie Wiles, Treasury Secretary Scott Bessent and National Cyber Director Sean Cairncross attended the sitdown.

According to two people familiar with the Friday meeting, granted anonymity to discuss sensitive talks, the conversation focused on Mythos, not the broader fight between the company and the administration. But both sides subsequently issued optimistic statements.

“There are definitely positive signs coming out of the White House and from the president stemming from [that] meeting,” said Paul Lekas, head of global public policy and government affairs at the Software Information Industry Association, a group that includes Anthropic among its members.

Asked about the apparent walkback, a White House spokesperson, who declined to comment on the record, pointed to remarks Trump made Tuesday on CNBC, in which the president praised the intelligence of Anthropic executives whom he had previously denounced as “Leftwing nut jobs.”

“They came to the White House a few days ago, and we had some very good talks with them,” Trump told the network. “And I think they’re shaping up. They’re very smart. … I like high-IQ people, and they definitely have high IQs."

Asked if the administration would make a deal with Anthropic, Trump said: “It’s possible.”

An Anthropic spokesperson declined to comment.

Trump’s CNBC comments were a “good sign” for the company, venture capitalist and White House technology adviser David Sacks said later Tuesday during an interview hosted by the Council on Foreign Relations. Sacks, who now serves as co-chair of the President’s Council of Advisors on Science and Technology, has been a relentless critic of Anthropic.

Saif Khan, a former adviser on emerging technology in the Biden administration and a fellow at the Institute for Progress think tank, agreed that people in the Trump administration appear “interested in trying to repair the relationship” with Anthropic. 

“Just hearing what the president said, I think that's definitely the direction it's trending in now,” said Khan. “I think they’ll come to a resolution."

Such an agreement would ease months of tensions, after Amodei insisted early this year that the company would not allow the Pentagon to use its Claude AI model to conduct mass surveillance of Americans or enable autonomous weapons. Trump and Hegseth accused Anthropic of trying to dictate government policy by imposing limits on what Hegseth framed as the “lawful use” of AI.

The fight between Washington and a leading American company sent a chill through the AI industry and threatened to undermine Trump’s quest for U.S. dominance in the technology, even as it opened new opportunities for Anthropic’s top corporate rival, OpenAI.

In federal court filings, the administration has also quietly walked back its most aggressive moves against Anthropic.

After the company sued the administration in March over its supply chain risk pronouncements, government lawyers argued that the label is “limited”: It only prohibited the Pentagon from working with the company, and barred Defense Department contractors from using Anthropic while working on government contracts.

That assurance ran counter to Hegseth’s previous statement on X that “effective immediately, no contractor, supplier, or partner that does business with the United States military may conduct any commercial activity with Anthropic.” Hegseth’s post was not a distinct or final agency action, the government’s lawyers wrote.

Trump had separately declared on Truth Social in February that federal agencies must “IMMEDIATELY CEASE all use of Anthropic’s technology,” though he gave them six months to begin phasing out the software. In spite of that supposed ban, offices such as the Commerce Department’s Center for AI Standards and Innovation are already testing Mythos, and other agencies like the Treasury Department have sought access.

Mythos has stirred widespread concern in the Trump administration, across the financial industry and in Europe because of its apparently advanced ability to find software flaws, which according to Anthropic presents a serious potential threat to any organization’s cybersecurity. Staff from at least two large federal agencies have reached out to Anthropic to express interest in integrating Claude Mythos into their cyber defense efforts, according to a former senior U.S. technology official with direct knowledge of the discussions.

Meanwhile, the Office of Management and Budget told government officials in an email last week, shared with POLITICO, that agencies may soon be allowed to use a "modified" version of Mythos.

Institute for Progress’ Khan said OMB’s guidance may rely on an injunction issued last month by a federal judge in northern California, which temporarily blocked both the government’s supply chain risk label against Anthropic and the president’s directive that agencies stop using its products.

“I think the California decision from a few weeks ago is binding here,” Khan said. “I think you saw right then — not even just last week, but well before that — that other agencies were starting to use Anthropic models again.”

But Lekas, from the software industry group, said the administration needs to mount a full-scale climbdown if it wants to reassure the tech industry that the Anthropic spat was an aberration and the government won’t again try to harm a top AI firm over a contract dispute. He said the administration cannot have it both ways — tacitly allowing federal agencies to use Mythos while preserving on paper the designation of Anthropic as a national security risk.

Another senior tech industry representative agreed that the technology lobby would need more clarity before it stopped pressuring the government on Anthropic.

Ball, the former Trump official, said the administration’s approach to Anthropic has been “absurd.” But while he agrees that the White House is now groping for a path forward on Mythos, he’s not sure that administration officials will be willing to completely abandon their animosity toward the company.

“We shouldn't let anyone pretend it's not a blaring contradiction, using this thing and also saying that it is the equivalent of a product made by a foreign enemy government,” said Ball. “But will they rectify the situation by walking back the policy? Very hard to say.”

John Hewitt Jones contributed to this report.