Pentagon Gives Anthropic Friday Ultimatum: Lift Claude AI Military Limits or Lose $200M Contract
The Pentagon told Anthropic to remove restrictions on its Claude model for military use by Friday, warning it could cancel a $200 million contract, designate the company a supply‑chain risk or invoke the Defense Production Act.
- Deadline and threat: The Pentagon set a Friday deadline for Anthropic to drop categorical limits on Claude or risk termination of a $200M contract (Fox News, Politico).
- Possible escalations: Officials said the department could label Anthropic a supply‑chain risk or use the Defense Production Act to compel access (Politico).
- Operational concern: Claude is currently the only advanced commercial model operating inside Pentagon classified networks under the agreement, raising immediate readiness stakes (Fox News).
Background and trigger
Anthropic has cultivated a reputation as a safety‑minded AI firm, imposing internal rules it says prevent misuse — notably refusing to permit fully autonomous weapons or enable mass domestic surveillance. Pentagon officials, however, argued that private rules should not block lawful military use.
The dispute escalated after Pentagon officials said Anthropic asked whether Claude had been used in the January 2026 U.S. operation that captured Venezuelan leader Nicolás Maduro. That question was viewed by officials as a potential signal the company might veto certain lawful uses; the reporting is based on multiple outlets and sources familiar with the discussions (Fox News, Politico).
Meeting and ultimatum
On Feb. 24, War Secretary Pete Hegseth met with Anthropic CEO Dario Amodei at the Pentagon. Sources say Hegseth praised Claude’s capabilities but delivered a firm message: the department cannot rely on a vendor that keeps categorical restrictions on lawful military use.
Hegseth demanded access for all lawful military uses without company approval or oversight and warned of three primary actions if Anthropic refused:
- Terminate the $200M contract that allows Claude inside classified networks (Fox News).
- Designate Anthropic a supply‑chain risk, a step that would push Pentagon partners away from the company (Politico).
- Invoke the Defense Production Act to compel access to needed technology during national security needs (Politico).
A source said Hegseth compared being restricted from lawful uses to telling the military it could not use a particular aircraft for a mission.
Public video of the meeting and exchanges has also circulated (meeting video).
Anthropic’s position
Anthropic says its policies are intended to reduce misuse risks as AI grows more powerful. The company denies it asked about specific operations in a way that would block lawful action and maintains its red lines bar fully autonomous weapon systems and mass surveillance of Americans. Company spokesperson Maya Humes said CEO Dario Amodei appreciated the Pentagon’s work and described the talks as “good‑faith conversations” about usage policy (Fox News, Politico).
Pentagon rationale and comparisons
Pentagon officials emphasize the need for predictable, unfettered access when integrating advanced tools into operations. They argue that categorical vendor restrictions could hamper urgent national security needs or slow battlefield decisions.
Officials noted other firms that have reportedly agreed to the department’s terms; for example, reporting indicates Elon Musk’s Grok accepted terms for classified integration, and that talks with Google and OpenAI are ongoing. The department told reporters it cannot build core capabilities on vendors that may decline lawful missions (Fox News, video).
Legal tools: supply chain risk and Defense Production Act
Designating a company a supply‑chain risk typically warns partners and can block participation in government contracts. Sources told reporters the Pentagon could use that label to push contractors away from Anthropic, even though the company is U.S.‑based.
The Defense Production Act gives the government authority to require production or access to critical materials and technologies during national emergencies — a mechanism the Pentagon said it could consider if necessary (Politico).
Claude’s role and stakes
Claude is currently the only advanced commercial AI model operating inside the Pentagon’s classified networks under the $200M contract awarded in summer 2025. Losing Claude would disrupt workflows and force the department to scramble for alternatives or accelerate classified access deals with other firms (Fox News).
Broader implications for AI guardrails
This dispute is an early test of who sets guardrails for advanced AI used in defense — private companies establishing safety rules or the Pentagon enforcing operational freedom. The outcome could shape future commercial‑defense partnerships as the military integrates AI into intelligence, logistics, planning and combat support (Politico).
Implications for Alamogordo, New Mexico
Economic impact: Alamogordo lies near Holloman AFB and testing ranges. Changes in Pentagon procurement for classified AI tools could affect local contractors who support those installations. If Anthropic is cut, regional firms could see contract shifts; conversely, companies that accept Pentagon terms might gain work.
Political consequences: Local leaders and conservative voters who prioritize defense may view strict Pentagon access rules as protecting readiness. If vendor replacement is slow, officials may press for accelerated solutions to avoid readiness gaps.
Social effects: Military families and civilian workers tied to defense installations could face hiring and training shifts. Sudden contractor changes can mean uncertainty in employment and base services.
Practical applications: Local contractors and small businesses should monitor procurement decisions. Firms that can meet requirements for unrestricted lawful use may win contracts. Schools and training programs might see increased demand for AI and cybersecurity skills.
Sources
Reporting in this article draws on multiple sources familiar with Pentagon‑Anthropic discussions and public reporting: Fox News (Pentagon gives AI firm ultimatum), Politico (Hegseth sets Friday deadline), and related public video of the meeting (YouTube).
(Reporting compiled from public reporting and sources familiar with the discussions. No Pentagon comment was available at time of publication.)
