Pentagon Demands Anthropic Lift AI Restrictions or Face Contract Termination

The U.S. Defense Department has issued a deadline of Friday to artificial intelligence company Anthropic, demanding that it remove restrictions on military use of its AI technology or face the termination of its $200 million government contract. The dispute centers on Anthropic's stated ethical boundaries regarding autonomous weapons and domestic surveillance, which the Pentagon argues limit lawful military applications.

Core Dispute and Deadline

Defense Secretary Pete Hegseth has informed Anthropic CEO Dario Amodei that the company must agree to allow the U.S. government to use its AI for all "lawful" purposes by Friday, or its contract, valued at up to $200 million, will be terminated.

This directive was conveyed during a meeting on Tuesday. Anthropic is currently the only major AI company that has not supplied its technology for the new U.S. military internal network, GenAI.mil, without specific usage limitations.

Anthropic's Stance

Anthropic CEO Dario Amodei has consistently articulated ethical concerns regarding the unrestricted deployment of AI by government entities. The company has drawn two specific lines: it will not permit its AI to be used for fully autonomous military targeting operations (AI-controlled weapons) or for domestic surveillance of U.S. citizens.

Amodei has previously written about the potential for powerful AI to monitor public sentiment and identify dissent.

Pentagon's Position

Defense officials contend that military operations necessitate AI tools free of built-in limitations. They assert that the military would be responsible for ensuring the legal application of Anthropic’s tools and that all orders issued by the Pentagon have been lawful.

Secretary Hegseth has previously stated that military AI systems must operate "without ideological constraints that limit lawful military applications" and that the Pentagon’s "AI will not be woke."

This term, "woke AI," has been used by Trump administration officials to characterize what they describe as strict safety protections for AI tools and perceived liberal biases in chatbots.

Potential Consequences

Should Anthropic not comply, the Defense Department could designate the company as a "supply chain risk." Additionally, the Pentagon has indicated the potential use of the Defense Production Act.

This 1950s law, typically invoked during national emergencies, grants the military authority to compel companies to produce or allow use of products deemed critical for national security, potentially overriding the company's explicit approval of how its products are used.

Broader Industry Context

Last summer, the Pentagon awarded defense contracts worth up to $200 million each to four AI companies: Anthropic, Google, OpenAI, and Elon Musk’s xAI. Initially, Anthropic was the first AI company approved for classified military networks, based on officials' assessment of its models as advanced and secure for sensitive applications. The other three companies initially operated in unclassified environments.

In contrast to Anthropic's stance, other contracted AI companies, including OpenAI, Google, and xAI, have agreed to permit the use of their AI tools in "lawful" military scenarios. xAI recently received approval for use in classified settings, and OpenAI announced its participation in the military’s secure AI platform for unclassified tasks in February.

Anthropic was founded in 2021 by individuals who left OpenAI, positioning itself as an AI company focused on safety. The company has aligned with the Biden administration on AI safeguards and has had prior disagreements with the Trump administration regarding AI export controls and regulation.

Calls for Oversight

Amos Toh, senior counsel at the Brennan Center’s Liberty and National Security Program at New York University, has called for increased congressional oversight or regulation of AI.

He specifically highlighted concerns regarding AI's potential use for surveillance of Americans, noting that existing legal frameworks may not be keeping pace with technological advancements.