Pentagon Ultimatum to Anthropic: Lift AI Restrictions or Lose Deal

WASHINGTON — The Pentagon has given AI firm Anthropic a Friday deadline to remove military-use restrictions on its Claude model, warning of contract cancellation, supply‑chain penalties or invoking the Defense Production Act if the company refuses to lift its guardrails.

• Deadline and stakes: Defense Secretary Pete Hegseth ordered Anthropic to drop categorical limits on Claude for lawful military use or face possible contract cancellation, supply‑chain risk designation, or Defense Production Act action.
• Operational reliance: Claude is reportedly the only commercial advanced model cleared for classified Pentagon networks under a roughly $200 million contract awarded in 2025, increasing leverage and risk.
• Company safety stance: Anthropic says its red lines bar fully autonomous weapons and mass domestic surveillance and that it did not seek to block specific operations; it frames limits as safety measures.
• Local impact: Termination or restrictions could ripple to communities such as Hesperia, California, affecting contractors, small businesses, and political debates over security versus civil‑liberties concerns.

The core dispute pits the Pentagon against Anthropic over who controls guardrails on advanced AI integrated into U.S. defense systems. Pentagon officials say a private company cannot place categorical vetoes on lawful military uses; Anthropic says defined limits are necessary to reduce misuse as models grow more powerful.

Reporting sources: Politico, Fox News, Responsible Statecraft, TechCrunch.

Defense leaders argue the military must be able to employ tools for all lawful purposes without a private vendor effectively reviewing or vetoing missions. Anthropic counters that some uses — notably fully autonomous weaponization and large‑scale domestic surveillance — present outsized risks and merit categorical limits.

Policy tension: The clash highlights a broader question about civilian contractors’ role in limiting government use of dual‑use technologies.

Officials say the dispute escalated after Anthropic reportedly asked whether Claude was used in a January operation to capture Venezuelan leader Nicolás Maduro. The Pentagon interpreted that question as implying Anthropic might seek approval authority over specific lawful missions. That exchange reportedly involved Palantir and raised alarm within the Defense Department.

Sources reporting the episode include Responsible Statecraft and Fox News.

According to people familiar with the talks, Secretary Hegseth warned Anthropic CEO Dario Amodei the department could take three major actions if the company refuses to drop its restrictions:

• Cancel the contract: Terminate the roughly $200 million agreement that allows Claude to operate on classified networks. (Fox News)
• Supply‑chain risk designation: Label Anthropic a supply chain risk, potentially blocking defense partnerships and contractor access. (Politico)
• Invoke the Defense Production Act: Compel access to critical technology if needed for national defense. (Politico)

“It would be like telling the military it could not use a particular aircraft for a mission,” officials say Hegseth framed the issue.

Anthropic denies seeking to vet or block specific operations. The company emphasizes safety‑first policies that bar fully autonomous weapons that select and fire on targets without meaningful human control and prohibit mass domestic surveillance of Americans. Anthropic representatives say these limits are intended to mitigate misuse as models scale.

Reporting on Anthropic’s statements: Fox News, TechCrunch.

Claude is reported to be the only advanced commercial model currently operating inside the Pentagon’s classified networks under the roughly $200 million contract awarded in 2025, making Anthropic cooperation critical and the department’s position more aggressive. Pentagon officials say they lack a true backup, though talks with other vendors — including xAI’s Grok — are underway.

The outcome could set precedent for future Pentagon partnerships and how the government balances reliance on private frontier AI firms with national security imperatives. (Responsible Statecraft, TechCrunch.)

In mid‑2025, OpenAI, Anthropic, Google, and xAI each won roughly $200 million Department of Defense contracts to prototype frontier AI tools for national security. Of those, Anthropic was the first approved for classified use. Responsible Statecraft reports xAI has agreed to allow its model to be used for any lawful DoD purpose, underscoring differing company positions on usage restrictions.

Economic impact: Hesperia hosts many service workers, small businesses, and residents tied to defense and tech supply chains. Contract termination or a supply‑chain designation could reduce work for local suppliers and ripple through contractors serving nearby military installations. (Responsible Statecraft)

Political consequences: Conservative‑leaning voters in Hesperia may support the Pentagon’s stance emphasizing civilian control of defense capabilities; local leaders could press for federal actions ensuring defense access to critical technologies. (Politico)

Social effects: Residents worried about domestic surveillance may back Anthropic’s red lines, while others prioritize security and support unfettered military access. The debate may surface in town halls and community discussions. (Fox News)

Practical notes: If the Defense Production Act is invoked or other federal measures secure broad access, supply chains and contracting opportunities may shift — with near‑term uncertainty for small businesses and potential federal investment favoring firms that accept DoD conditions. (Politico)

This story is based on reporting and sources familiar with the discussions and negotiations. Major outlets covering the developments include Politico, Fox News, Responsible Statecraft, and TechCrunch. The dispute remained fluid as Friday’s deadline approached.