WASHINGTON, February 26, 2026 : The U.S. Department of Defense has issued a formal “best and final offer” to artificial intelligence company Anthropic, requiring the firm to grant the military full lawful access to its Claude AI model without corporate-imposed usage restrictions. The proposal was delivered on the evening of February 25 and carries a deadline of Friday, February 27, 2026, at 5:01 p.m., as set by Defense Secretary Pete Hegseth.
Under the terms outlined by the Pentagon, Anthropic must permit the use of Claude for all lawful military purposes. Failure to accept the conditions would result in the termination of its existing $200 million defense contract and could trigger additional federal actions, including designation as a supply chain risk, effectively blacklisting the company from U.S. defense supply chains, and potential invocation of the Defense Production Act.
Contract Background and Classified Deployment
Anthropic was awarded a two-year prototype Other Transaction Agreement valued at up to $200 million in July 2025 through the Department of Defense’s Chief Digital and Artificial Intelligence Office (CDAO). The agreement supports the development and deployment of frontier AI capabilities to advance U.S. national security objectives.
Claude became the first AI model approved for operation on classified U.S. military networks and remains the only AI model currently operational for sensitive classified military work. Through a strategic partnership with Palantir Technologies, Anthropic secured direct integration within sensitive Pentagon systems. At present, Anthropic remains the only AI developer with an operational model on classified Department of Defense networks.
The Pentagon simultaneously awarded similar contracts of up to $200 million each to OpenAI, Google DeepMind, and xAI in July 2025 as part of a diversified AI procurement strategy. However, those systems had not matched Claude’s classified deployment status at the time of initial integration.
Core Dispute Over Usage Restrictions
The dispute between the Pentagon and Anthropic centers on the company’s terms of service and built-in guardrails governing military applications of Claude.
Anthropic, led by CEO Dario Amodei, has insisted on maintaining guardrails for military use, including prohibitions on domestic mass surveillance operations and fully autonomous lethal weapons systems. The company has also stipulated that its model must not make final lethal targeting decisions without meaningful human oversight, citing risks associated with AI-generated inaccuracies, commonly referred to as hallucinations, and the potential for operational errors.
During negotiations in December 2025, Anthropic agreed to permit Claude’s use for missile defense and cyber defense applications. However, the Department of Defense is seeking broader authorization without vendor-imposed limitations beyond compliance with U.S. law.
Defense officials have stated that the standard governing military AI procurement is “all lawful use,” arguing that determinations regarding legality in armed conflict and national security operations fall under federal jurisdiction rather than private corporate policy. Secretary Hegseth has publicly stated that the Department will not procure systems constrained by what officials describe as corporate ideological limitations and has emphasized the need for AI systems capable of full lawful military application, particularly in time-sensitive national security scenarios such as missile defense operations.
Federal Enforcement Measures Under Consideration
If Anthropic declines the final offer by the established deadline, the Pentagon has indicated it will immediately terminate the July 2025 contract.
Additionally, the Department is prepared to designate Anthropic as a “supply chain risk,” a classification that would require U.S. defense contractors to remove Anthropic software from their operational networks and discontinue integration into defense systems. Officials have indicated that such a step would effectively bar the company from participating in future defense-related procurements.
Officials have also stated that the Department of Defense is considering invoking the Defense Production Act. This federal statute would grant the government authority to compel compliance with national security requirements, potentially overriding Anthropic’s corporate usage restrictions if deemed necessary in the interest of defense readiness.
xAI Signs Pentagon Deal; Grok Entering Classified Systems
Amid the ongoing dispute, xAI has signed a Pentagon agreement enabling its Grok model to enter classified U.S. military systems. Defense officials confirmed that the deal has been finalized, positioning Grok as an alternative frontier AI system for sensitive national security applications.
The agreement follows internal approvals for classified deployment and expands the Department’s access to AI systems beyond Claude. The Pentagon has stated that multiple vendors are necessary to ensure redundancy and operational flexibility across intelligence, cyber, and missile defense domains.
Recent Escalation and Operational Context
The dispute intensified following reports that Claude was integrated into a January 2026 military operation that resulted in the capture of former Venezuelan President Nicolás Maduro. Details regarding Claude’s specific operational role have not been publicly disclosed.
Defense Secretary Hegseth met with CEO Dario Amodei on February 24, 2026, to discuss the matter prior to the issuance of the formal proposal.
As of February 26, 2026, Anthropic has not publicly responded to the Pentagon’s final offer.
Broader AI Integration Strategy
The situation forms part of the Department of Defense’s broader initiative to integrate frontier artificial intelligence systems across operational, intelligence, cyber, and missile defense domains. The July 2025 contract awards to Anthropic, OpenAI, Google DeepMind, and xAI reflect a diversified procurement strategy aimed at accelerating AI adoption within the military while maintaining competitive development pathways.
The outcome of the February 27 deadline may influence future federal procurement standards governing the interaction between commercial AI safety policies and national defense requirements, particularly as additional AI systems, including Grok, transition into classified operational environments.
——— End of Article ———
