Anthropic’s $200 million Pentagon contract collapsed on February 27, 2026, after the White House directed all federal agencies to stop using the company’s Claude AI models. The breakdown followed Anthropic’s refusal to remove two usage restrictions the Pentagon had originally agreed to when the contract was signed in 2024: prohibitions on using Claude for mass surveillance of American citizens and for fully autonomous lethal weapons systems.
Secretary Pete Hegseth moved swiftly, directing the Department of War to designate Anthropic a “Supply-Chain Risk to National Security” — a classification historically applied to foreign technology companies such as Huawei and Kaspersky Lab. The designation gives the Pentagon a 180-day window to remove Claude from its systems and effectively orders contractors and partners to halt commercial activity with Anthropic immediately.
The central dispute is straightforward. The Pentagon sought unrestricted access to Claude for any use it deemed legally permissible — a standard the department calls “all lawful use.” Anthropic CEO Dario Amodei declined, describing the two prohibited applications as firm red lines and arguing that mass domestic surveillance is incompatible with democratic values, and that fully autonomous weapons systems risk unintended escalation or operational failure. Hegseth characterized the refusal as arrogance. Anthropic characterized the guardrails as necessary. Neither side moved.
The commercial picture for Anthropic remains strong despite the rupture. The company’s Claude Code service has grown into a division generating over $2.5 billion in annual recurring revenue less than a year after launch. Earlier in February, Anthropic closed a $30 billion Series G financing round at a reported $380 billion valuation. Enterprise clients including Salesforce, Spotify, Novo Nordisk, and Thomson Reuters have reported meaningful productivity gains from Claude deployments. In the consumer market, the Claude app climbed to the second most downloaded application in the Apple App Store in the immediate aftermath of the dispute, reflecting a wave of support from developers, technologists, and general users.
Anthropic’s competitors wasted little time. OpenAI announced a Pentagon deal accompanied by what the company described as two safety principles, though the specific contractual language remains unclear and may differ materially from Anthropic’s red lines. OpenAI also disclosed a $110 billion investment round led by Amazon, Nvidia, and SoftBank. Elon Musk’s xAI separately confirmed that its Grok model had been cleared for use in classified government systems under the all-lawful-use standard, though early feedback from government and military personnel reportedly places Grok below expectations.
Anthropic has stated it intends to challenge the national security designation in court and has advised its commercial customers to continue using its products, with the explicit exception of military applications.
For enterprise technology leaders, the episode raises a practical question that has nothing to do with the politics surrounding it. A single executive decision — in this case, a refusal to modify usage policies — produced an immediate federal blacklisting that cascaded into contractor obligations and procurement restrictions within hours. Any enterprise that has built critical workflows or infrastructure on a single AI provider now has concrete evidence of the concentration risk that entails. The Anthropic situation did not create that risk. It simply made it visible. Vendor diversification, contractual clarity around usage policies, and documented contingency planning for provider disruption are no longer theoretical governance concerns. They are operational necessities demonstrated by an event that moved from contract dispute to national security designation in a single news cycle.
Source: Original reporting
