Anthropic CEO Dario Amodei is set to meet with Defense Secretary Pete Hegseth at the Pentagon on Tuesday, February 24, 2026, to discuss how the U.S. military could use the startup’s AI models, according to a senior Department of Defense official. Axios first reported the planned meeting.
The sit-down comes as negotiations between Anthropic and the DoD have reportedly hit a rough patch over how the company’s technology can be used. Anthropic wants assurances its models won’t be used for autonomous weapons or domestic surveillance, while the Pentagon’s position has been that it wants access “for all lawful use cases,” without added limitations.
Why Anthropic and the Pentagon are at odds
Reporting suggests the dispute isn’t about whether Anthropic’s models can support national security work, but about guardrails. Anthropic has publicly positioned itself as a “safety-first” AI lab, and the company is seeking clearer commitments that its models won’t be applied to autonomous weapons systems or used to spy on Americans.
The Pentagon, by contrast, has signaled it wants flexibility to deploy the models across “all lawful” uses—language that can cover a wide range of military and intelligence workflows. That difference in framing has become the key sticking point.
Anthropic’s current foothold in national security
Anthropic has been aggressively building a government-facing business around Claude. The company’s Claude Gov offering has been described as a tailored set of models built for U.S. national security customers and deployment in sensitive environments.
However, it’s worth tightening the common “only company on classified networks” claim: Anthropic is widely reported as an early mover here, but the DoD has also been pushing other major AI vendors to expand availability on classified systems—so the exact “only” status can depend on timing and how “deployed” is defined.
The $200 million DoD deal and what it means
Anthropic’s relationship with the DoD isn’t hypothetical. In July 2025, the company announced a two-year prototype Other Transaction Agreement (OTA) with a $200 million ceiling through the DoD’s Chief Digital and AI Office (CDAO). The goal: develop and test responsible AI capabilities that can be used in defense operations.
If current talks stall, it could complicate how quickly Anthropic expands those deployments—or whether the Pentagon shifts more of its attention to other providers willing to accept broader usage terms.
Political pressure is rising
The clash also lands in a tense political moment. The Trump administration has been publicly critical of Anthropic in recent months, and the Pentagon’s posture—pushing for broad “lawful use” language—reflects a broader trend: national security agencies want access to frontier models, but they don’t want private vendors setting policy-like constraints.
Tuesday’s meeting could be a chance to find a middle ground: language strict enough for Anthropic’s safety commitments, but flexible enough for the Pentagon’s mission demands.
Anthropic’s momentum outside Washington
The timing is notable because Anthropic has been on a tear commercially. The company, founded in 2021 by former OpenAI researchers and best known for the Claude model family, announced earlier this month that it raised $30 billion in funding, putting its post-money valuation at $380 billion.
That kind of scale changes the stakes. Anthropic doesn’t need Washington to survive—but access to national security workloads can be a strategic wedge into regulated industries, long-term contracts, and high-value compute demand. Whether it can keep that wedge without compromising its guardrails is the question now sitting on the table.