Anthropic opens Claude to civilian US agencies while Pentagon fight drags on

Anthropic is now selling Claude for Government to US federal and state agencies, running in a FedRAMP High environment — the strictest US security level for cloud services. The Pentagon won't be a customer. The Department of Defense still classifies Anthropic as a supply chain risk, according to The Decoder's report on the launch, and that classification remains in place even as the company extends its reach across the rest of government.

The move marks a significant expansion of Anthropic's public-sector business into the civilian side of the US government, a market where rivals including Microsoft, Google, and OpenAI have already built footholds through long-standing cloud contracts and, in OpenAI's case, a fast-growing government offering of its own.

What agencies get

Claude for Government gives agencies the same feature set Anthropic sells to enterprise customers, according to the announcement. That parity matters for government buyers who have historically faced stripped-down or delayed versions of commercial AI tools. Federal and state agencies adopting the platform will work with the current generation of Claude models inside an environment cleared for some of the government's most sensitive unclassified workloads.

The commercial terms are structured around usage-based pricing with fixed spending caps. Agencies can also set budgets for each department — a control designed to address one of the biggest practical concerns in government AI adoption: unpredictable costs. Procurement officers under scrutiny over AI spending can bind individual bureaus to hard limits rather than tracking consumption after the fact.

FedRAMP High authorization is the key technical detail. The Federal Risk and Authorization Management Program's "High" baseline applies to cloud services handling the government's most sensitive unclassified data, and achieving it requires an extensive security review process. By operating at that level from launch, Anthropic positions Claude for Government for work well beyond pilot projects — regulatory drafting, analysis of sensitive administrative records, and large-scale document processing across civilian departments.

The Pentagon problem

The launch's most notable exclusion is the Department of Defense. The Pentagon will not use the new platform because it still classifies Anthropic as a supply chain risk, The Decoder reports. That designation — the same category of concern US officials have applied to certain foreign technology vendors — keeps Anthropic out of defense procurement even as civilian agencies gain access to its models at the highest cloud security tier.

The contradiction is stark. An AI company can simultaneously meet the strictest civilian security requirements in the federal cloud ecosystem and fail defense-side vetting. For Anthropic, the gap represents a lost revenue channel at a moment when defense contracts have become a major growth engine for AI vendors across the industry. Competitors have moved aggressively into national security work, and each month the supply chain classification stands, Anthropic cedes ground in a segment many investors now treat as a baseline expectation for frontier-model companies.

The report frames the situation as a fight that "drags on" — language suggesting an unresolved dispute rather than a pending formality. How and when Anthropic might clear the Pentagon's bar remains unclear from the available information.

Why it matters

The US government has become one of the most contested markets in AI. Federal and state agencies control multi-billion-dollar technology budgets, and their choices of model providers shape which companies accumulate the revenue, deployment data, and regulatory credibility needed to compete at the frontier. A FedRAMP High offering puts Anthropic in the running for that spend across the civilian government — the larger share of federal IT procurement by dollar volume.

The launch also signals how AI vendors are segmenting the public sector. Civilian-first entry — with enterprise-equivalent features, usage-based pricing, and departmental budget controls — lets a company build government references and compliance track record while defense-related issues are resolved separately. For agency customers, the spending caps and per-department budgets address procurement pain points that have slowed earlier AI rollouts in government.

Anthropic's next test is whether civilian adoption translates into contracts at scale — and whether the Pentagon's supply chain classification changes before the defense AI market consolidates around its competitors.