A significant standoff has emerged between artificial intelligence company Anthropic and the U.S. Pentagon regarding the terms of use for Anthropic's advanced AI technology, specifically its chatbot Claude. Anthropic CEO Dario Amodei publicly stated Thursday that the company "cannot in good conscience accede" to the Pentagon's demands for wider application of its technology. The core of Anthropic's refusal stems from deep concerns that the proposed contract language still makes "virtually no progress on preventing Claude's use for mass surveillance of Americans or in fully autonomous weapons." These are applications explicitly prohibited by Anthropic's internal policies for its AI models.

This dispute, which has been escalating for months, comes to a head with a Friday deadline imposed by the Defense Department. The Pentagon's top spokesman, Sean Parnell, responded to Amodei's comments, reiterating the military's intent to use Anthropic’s AI technology only in "lawful ways." Parnell asserted on social media that the Pentagon "has no interest in using AI to conduct mass surveillance of Americans (which is illegal) nor do we want to use AI to develop autonomous weapons that operate without human involvement." He emphasized that the Pentagon wants to "use Anthropic’s model for all lawful purposes" and will not allow any company to "dictate the terms regarding how we make operational decisions," suggesting that limiting use could jeopardize critical military operations.

During a high-stakes meeting on Tuesday between Defense Secretary Pete Hegseth and Amodei, military officials reportedly warned Anthropic of severe repercussions should they fail to comply. These threats included the cancellation of Anthropic's contract, designating the company as a supply chain risk, or even invoking the Cold War-era Defense Production Act. This act would grant the military sweeping authority to utilize Anthropic's products irrespective of the company's approval. Amodei, however, highlighted the inherent contradiction in these threats, noting that "one labels us a security risk; the other labels Claude as essential to national security." Parnell's subsequent Thursday post on X omitted the threat of the Defense Production Act, focusing instead on the Friday deadline: "Otherwise, we will terminate our partnership with Anthropic and deem them a supply chain risk."

Anthropic is notably the last among its major AI peers—including Google, OpenAI, and Elon Musk’s xAI—to resist supplying its technology to a new internal network utilized by the U.S. military. Amodei expressed hope that the Pentagon would reconsider its demands, citing "the substantial value that Anthropic’s technology provides to our armed forces." Should an agreement not be reached, Anthropic has stated its readiness to facilitate a smooth transition to another provider.

The public nature of this disagreement has drawn criticism from Capitol Hill. Senator Thom Tillis, a Republican from North Carolina, criticized the Pentagon's handling of the matter as "unprofessional," arguing that Anthropic is "trying to do their best to help us from ourselves." Tillis questioned the public discussion of such sensitive negotiations, advocating for private dialogue when a vendor resists market opportunities due to potential negative consequences. Similarly, Senator Mark Warner of Virginia, the ranking Democrat on the Senate Intelligence Committee, expressed his "deeply disturbed" by reports of the Pentagon "working to bully a leading U.S. company." Warner described this as further evidence that the Department of Defense is ignoring AI governance, underscoring the urgent need for Congress to establish strong, binding AI governance mechanisms specifically for national security contexts. This aligns with earlier statements from Defense Secretary Hegseth, who in February suggested that lawyers should provide "sound constitutional advice" rather than acting as "roadblocks" to military operations.