Bold stance, tough truths: Anthropic says it cannot in good conscience grant unrestricted access to its AI, even as the Pentagon presses for a broader, unfettered military use of Claude.
Anthropic CEO Dario Amodei stated on Thursday that the company cannot agree to the Defense Department’s latest contract language, which he argues would largely fail to stop Claude from being used for mass surveillance of Americans or for fully autonomous weapons. While Anthropic isn’t walking away from negotiations entirely, the new terms appear to stall progress on limiting those risks.
The Pentagon’s top spokesperson, Sean Parnell, emphasized that the military has no interest in enabling mass surveillance or autonomous weapons that operate without human input. Anthropic’s policies already block such use cases, but the clash marks a rare public feud with a government client that also includes Google, OpenAI, and Elon Musk’s xAI in separate contracts.
Anthropic’s position mirrors a broader stance: the Department should choose contractors who align with its strategic vision. Amodei expressed optimism about the value Claude brings to U.S. armed forces and urged the administration to reconsider.
Defense Secretary Pete Hegseth issued an ultimatum earlier in the week: open Anthropic’s technology for unrestricted military use by Friday or risk losing the contract. Officials warned of potential escalations, including labeling the company a supply chain risk or invoking the Defense Production Act to grant the military broader access to its products.
Amodei countered that the two threats—being deemed a security risk and being deemed essential to national security—are inherently contradictory.
Parnell reiterated that the Pentagon wants to use Anthropic’s model for all lawful purposes, but did not provide specifics. He argued that expanding use would help avoid jeopardizing critical military operations and stressed that no company should dictate how the military makes decisions.
Meanwhile, talks have been ongoing for months. Amodei acknowledged the substantial value of Anthropic’s technology for the armed forces and expressed hope for a reconsideration. If negotiations fail, he indicated Anthropic would assist in a smooth transition to another provider.
Politicians weighed in. Sen. Thom Tillis criticized the public nature of the dispute, calling it unprofessional and suggesting that sensitive vendor discussions should occur behind closed doors. He urged listening to the vendor’s concerns before pursuing a public tug-of-war.
Sen. Mark Warner expressed concern that the Pentagon may be bullying a major U.S. company and framed the episode as evidence that stronger, binding AI governance for national security contexts is needed in Congress.
All parties acknowledge a culture shift within the defense legal community, with some leadership moves at the Pentagon raising questions about how constitutional and legal advice is weighed in war-fighting decisions. The episode underscores the tension between rapid AI advancement and the safeguards expected in legal and ethical governance.
Note: This summary reflects ongoing reporting and does not modify the substance of the original reporting.”}
