A deep ideological and legal rift between the Pentagon and Anthropic has not stopped the National Security Agency (NSA) from deploying the startup’s most advanced artificial intelligence (AI).
The NSA is reportedly using Mythos Preview, a highly restricted model capable of sophisticated offensive cyber operations, even as the Department of Defense continues to label Anthropic a “supply chain risk,” according to an Axios report citing people familiar with the matter.
The internal contradiction highlights a growing tension within the federal government, where the urgent need for elite cybersecurity tools is clashing with a messy, high stakes battle over how military AI should be governed.
The friction reached a breaking point in February, when the Pentagon moved to sever ties with Anthropic, attempting to force its various vendors to follow suit. That legal battle remains ongoing, yet the military’s actual usage of Anthropic’s technology appears to be expanding rather than contracting.
Although the specific nature of the NSA’s work with Mythos remains classified, similar organizations granted access to the model use it primarily to scan their internal environments for exploitable vulnerabilities. The model is so potent that Anthropic restricted its release to about 40 vetted organizations, fearing its offensive capabilities could be catastrophic if released to the public.
This paradox of the Pentagon arguing in court that Anthropic threatens national security while its own agencies utilize the company’s “dangerous” elite tools stems from a breakdown in contract negotiations earlier this year.
During those talks, the Department of Defense demanded that Anthropic’s Claude model be made available for “all lawful purposes.”
Anthropic leadership pushed back, insisting on strict guardrails that would prohibit the technology from being used for mass domestic surveillance or the development of autonomous lethal weaponry. Some defense officials viewed this refusal as proof that the company is an unreliable partner; others in the administration are reportedly eager to end the feud to ensure the U.S. maintains a technological edge.
Efforts to bridge this gap reached the highest levels of government last Friday, when Anthropic CEO Dario Amodei met with White House Chief of Staff Susie Wiles and Treasury Secretary Scott Bessent.
The meeting, described by participants as productive, focused on Anthropic’s security practices and how other civilian departments might engage with Mythos.
Anthropic and the Pentagon declined to comment on the record. However, the high-level diplomatic outreach suggests the White House is seeking a path forward that bypasses the military’s current gridlock.
“They came to the White House a few days ago, and we had some very good talks with them,” Trump told CNBC’s Squawk Box on Tuesday. “And I think they’re shaping up. They’re very smart, and I think they can be of great use. I like smart people … I think we’ll get along with them just fine.”
For now, the NSA continues to leverage the very tools the Pentagon claims it cannot trust, illustrating a government caught between its security requirements and its regulatory principles.
