Anthropic’s updated cybersecurity framework might help restore its standing with the government.

For almost two months, the Trump administration has been at odds with AI firm Anthropic. The administration has labeled the company a “RADICAL LEFT, WOKE COMPANY” filled with “Leftwing nut jobs” and deemed it a threat to national security. However, there are reports indicating thawing tensions between the two sides, attributed to Anthropic’s newly launched cybersecurity-focused model: Claude Mythos Preview.

The relationship between Anthropic and the Pentagon soured rapidly in late February after the company declined to compromise on two crucial issues: utilizing its technology for domestic mass surveillance or developing fully autonomous lethal weapons without human oversight. Previously, Anthropic’s technology had been significantly employed by the DoD, and it was the first firm whose models received clearance to function on classified military networks. The impasse resulted in public clashes on social media, Anthropic being classified as a “supply chain risk,” the firm filing a lawsuit challenging that classification, and a temporary injunction lifting its ban.

Anthropic has recently endeavored to mend its relationship with the US government, at least partially, through the introduction of Mythos Preview. And according to reports indicating that CEO Dario Amodei participated in a meeting at the White House on Friday, it may be having an effect. Anthropic acknowledged the meeting on Friday. “Today, Anthropic’s CEO Dario Amodei engaged with senior administration members for a fruitful discussion regarding how Anthropic and the US government can cooperate on key mutual interests such as cybersecurity, maintaining America’s lead in the AI race, and AI safety,” stated Anthropic representative Max Young. “This meeting demonstrated Anthropic’s continual commitment to collaborating with the US government on the advancement of responsible AI. We appreciate their time and are eager to continue these dialogues.”

Mythos Preview was launched to great fanfare concerning its abilities — notably its capacity to identify security vulnerabilities across nearly all major web browsers and operating systems. Anthropic asserts that this model represents its most advanced iteration to date, currently offered exclusively for private access. It is positioned as a tool for highlighting critical vulnerabilities within widely utilized internet infrastructure, allowing companies like Apple, Nvidia, and JPMorgan Chase — all of which have already opted to employ it — to address these issues before they can be exploited by malicious actors. The rollout of Mythos Preview has purportedly already sparked emergency discussions among US banking executives and Federal Reserve Chairman Jerome Powell.

The Trump administration also appears to be paying attention. In a statement regarding Mythos Preview, Anthropic noted that it had been engaged in “ongoing conversations with US government representatives about Claude Mythos Preview and its offensive and defensive cybersecurity functionalities.” Earlier this month, when The Verge inquired, Dianne Penn, a product management leader at Anthropic, verified that the firm had “briefed senior officials in the US government about Mythos and its capabilities,” asserting that the company remains “dedicated to collaborating with various government levels.” The company opted not to specify exactly whom had been briefed.

Additionally, Anthropic has reportedly recently brought on Ballard Partners, a lobbying firm associated with Trump, prompting further speculation that a potential agreement between Anthropic and the White House might be forthcoming.

On Friday, Axios reported that Amodei was slated for a meeting with White House chief of staff Susie Wiles later that day. Describing the reasons for the meeting, a source familiar with the discussions stated, “It would be exceedingly irresponsible for the U.S. government to deprive itself of the technological advancements offered by the new model” and emphasized that “it would be a boon for China.” The report also mentioned that “parts of the U.S. intelligence community, along with the Cybersecurity and Infrastructure Security Agency (CISA, part of Homeland Security)” are testing Mythos Preview, while other departments and agencies are showing interest.

Should Amodei’s discussions lead to deeper integration of Anthropic’s Claude within governmental operations, it is conceivable that the DoD might revise its stance on Claude as well. This would mark a somewhat anticlimactic resolution to a contentious dispute surrounding national security — not the first instance of the administration abruptly altering its position.