Pentagon Pressures Anthropic To Lift AI Guardrails Or Risk $200M Contract

The Pentagon has drawn a hard line in the sand: lift the restrictions on your AI system or lose a $200 million defense contract.

Defense officials have reportedly given artificial intelligence firm Anthropic until Friday to remove limits on how its Claude AI model can be used by the U.S. military. If the company refuses, the Department of Defense is prepared to terminate the contract, designate the firm a supply chain risk, or even invoke the Defense Production Act to compel access.

The confrontation marks one of the first major power struggles over who controls advanced AI inside America’s national security infrastructure.

Claude is currently the only advanced commercial AI model operating inside the Pentagon’s classified networks under a contract awarded in summer 2025.

That alone raises the stakes. If the contract is canceled:

• Sensitive workflows would need to transition to a new provider

• Classified systems could face disruption

• The Pentagon would lose access to an already integrated AI platform

According to defense officials, the Department of Defense cannot depend on a contractor that maintains categorical restrictions on certain lawful military applications.

War Secretary Pete Hegseth reportedly told Anthropic CEO Dario Amodei that the military must be able to use AI tools for all lawful purposes without seeking corporate approval for specific missions.

The friction reportedly intensified after Pentagon officials believed Anthropic questioned whether Claude was being used in a January operation to capture Venezuelan leader Nicolás Maduro in a way that suggested possible disapproval.

Anthropic denies discussing specific operations and says its red lines focus on:

• Fully autonomous weapons

• Mass surveillance of Americans

The company has built its brand around AI safety and argues its guardrails are designed to prevent misuse as systems grow more powerful.

A senior Pentagon official countered that lawful military use already requires human oversight and compliance with U.S. law. “There’s always a human involved,” the official said.

The Pentagon is reportedly considering three significant pressure tactics:

1. Contract termination – Ending the $200 million agreement.

2. Supply chain risk designation – Potentially restricting Anthropic’s ability to work with federal contractors.

3. Defense Production Act invocation – Using national security authorities to compel access to critical technology.

The Defense Production Act has historically been used to prioritize industrial output in wartime or national emergencies. Applying it to frontier AI systems would signal how strategically vital such tools have become.

The dispute unfolds as the U.S. military accelerates AI integration across logistics, intelligence analysis, cybersecurity, and operational planning.

Global AI investment exceeded $150 billion last year, with defense applications representing a rapidly expanding segment. Meanwhile, China has openly declared its ambition to dominate AI by 2030, intensifying pressure on U.S. policymakers to avoid technological bottlenecks.

Pentagon officials have suggested that Elon Musk’s Grok AI has agreed to allow use for all lawful purposes, and that other firms are nearing similar arrangements.

That leaves Anthropic facing a consequential decision: maintain strict corporate guardrails or fully align with Pentagon doctrine.

At its core, the showdown is about authority.

Should private companies dictate ethical limits on battlefield technologies? Or does the Constitution vest that authority solely in civilian leadership overseeing the armed forces?

The answer could shape how advanced AI is deployed in national defense for years to come.

For continued coverage on AI, defense policy, and national security, share this article or subscribe to our newsletter for updates.