Anthropic And Trump: Is A Truce Near?

The Trump administration is publicly dialing back the heat in its standoff with the artificial intelligence developer Anthropic over ethical limits on the use of its AI — as worries about the company’s powerful new hacking tool undermine the president’s attempt to sever the startup’s ties to the government.

Anthropic CEO Dario Amodei’s high-stakes meeting Friday afternoon with White House chief of staff Susie Wiles, Treasury Secretary Scott Bessent and National Cyber Director Sean Cairncross was a contrast to the months of sniping and litigation between the two sides.

Any truce with Anthropic would be a dramatic climb-down for President Donald Trump and Defense Secretary Pete Hegseth, who had attempted to wield the federal government’s vast market power against one of the United States’ most advanced AI companies. Anthropic gained new leverage last week after announcing a potent new cybersecurity tool known as Mythos, prompting federal agencies to scramble to gain access to the software so they can evaluate how much danger it poses.

After Friday’s meeting, both Anthropic and White House spokespeople separately described the gathering as a “productive” starting point.

“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 in a statement. It added that the administration intends to invite other leading AI companies to the White House for similar discussions.

Anthropic said the meeting reflected the company’s “ongoing commitment to engaging with the U.S. government on the development of responsible AI” and included discussions around partnering on shared priorities including cybersecurity. The statement made no mention of Mythos.

The feud boiled over in February after Amodei insisted that Anthropic would not allow the Defense Department to use the company’s AI software to carry out mass surveillance of Americans or deploy fully autonomous weapons. That ran afoul of Hegseth’s insistence that the department be allowed to use the AI tools for any “legal” purposes, with the government having final say.

The administration then dropped two hammers: In late February, Trump ordered all federal agencies to stop using Anthropic, denouncing it as a “woke” company run by “leftwing nut jobs.” Days later, the Pentagon declared Anthropic a risk to the national security supply chain — a designation, normally reserved for companies tied to foreign adversaries, that could force any contractor doing business with the Defense Department to cease using Anthropic software.

The punishments threatened not only Anthropic’s $200 million Pentagon contract but the independence of the entire AI industry, legal experts told POLITICO at the time.

Anthropic had mixed success in challenging those penalties in court: A federal judge in California temporarily blocked the Pentagon’s supply chain risk label, but the D.C. Circuit Court of Appeals refused to take similar action this month.

Then came Mythos.

Cyber tool inspires transatlantic fears

As the advanced cybersecurity powers of the new AI model became apparent, federal agencies and government officials quietly sidestepped Trump’s ban on working with the startup, including the Commerce Department’s Center for AI Standards and Innovation. Bessent’s Treasury Department — which recently sought access to Mythos in order to determine its implications for banking firms’ cyberdefenses — is also among the agencies clamoring for guidance on how to proceed.

The software is also alarming regulators in Europe, who have told POLITICO they have not been able to gain access to Mythos.

U.S. government agency tech leaders sought access to the model after Anthropic earlier this year began testing the model and granted limited access to a select group of companies, including JPMorgan, Amazon and Apple.

Anthropic began limited sharing of Mythos after finding it had hacking capabilities far outstripping those of previous AI models. This includes the ability to autonomously identify and exploit complex software vulnerabilities, such as so-called zero-day flaws, which even some of the sharpest human minds are unable to patch.

The AI startup also wrote that the model could carry out end-to-end cyberattacks autonomously, including by navigating enterprise IT systems and chaining together exploits. It could also act as a force-multiplier for research needed to build chemical and biological weapons, and in certain instances, made efforts to cover its tracks when attacking systems, according to Anthropic’s report on the model’s capabilities and its safety assessments.

Those findings and others have inspired fears that the model could be co-opted to launch powerful cyberattacks with relative ease if it fell into the wrong hands. Logan Graham, a senior security researcher at Anthropic, previously told POLITICO that researchers and tech firms had been given early access to Mythos so they could find flaws in their critical code before state-backed hackers or cybercriminals could exploit them.

“Within six, 12 or 24 months, these kinds of capabilities could be just broadly available to everybody in the world,” Graham said.

Agencies want access

According to an email sent to federal agencies and obtained by POLITICO, the Office of Management and Budget is now mulling whether agencies will be allowed to use a “modified” version of Mythos.

“We’re working closely with model providers, other industry partners, and the intelligence community to ensure the appropriate guardrails and safeguards are in place before potentially releasing a modified version of the model to agencies,” Gregory Barbaccia, OMB’s chief information officer, wrote to multiple federal agencies earlier this week. Barbaccia added that OMB and the Office of the National Cyber Director will “continue to keep everyone updated and are expecting to have more information in the coming weeks.”

Cairncross, who heads the cyber office, convened a call this week with representatives from private sector companies to discuss Mythos, according to one participant on the call, granted anonymity because they were not authorized to speak publicly about it.

Cairncross said his office has the backing of Wiles and Vice President JD Vance to lead the Trump administration’s response to the hacking capabilities of Anthropic’s powerful new model, said the person. During the roughly 10-minute call, Cairncross both commended Mythos as a sign of U.S. AI innovation and warned companies about the risks it presented to their networks.

Spokespeople for OMB, the cyber office, Treasury and the vice president’s office did not respond to POLITICO’s requests for comment.

Last week, the heads of several major U.S. banks met with Bessent and Federal Reserve Chair Jerome Powell to discuss concerns about AI’s cyber capabilities and how to leverage that power to enhance critical infrastructure security.

After meeting with Bessent and other world financial leaders, Canadian Finance Minister François-Philippe Champagne told reporters on Friday that Mythos “requires all of our attention” to “maintain the integrity” of global financial institutions.

Bob King, John Hewitt Jones, Rosie Perper and Mickey Djuric contributed to this report.