Anthropic Defies Pentagon Ultimatum Over Unrestricted Military AI Use

The escalating confrontation between Anthropic and the U.S. Department of Defense represents a watershed moment for the artificial intelligence industry, testing the boundary between corporate ethics and national security imperatives. At the heart of the dispute is the Pentagon's demand for unrestricted use of Anthropic’s Claude models, a request that CEO Dario Amodei has characterized as a violation of the company’s "conscience." By refusing to comply with a February 27 deadline, Anthropic has positioned itself as the primary institutional holdout against the militarization of advanced AI, even as competitors like OpenAI and Google have historically navigated more complex, often quieter, relationships with defense agencies.

The Pentagon’s ultimatum is particularly aggressive, leveraging the Defense Production Act (DPA)—a Cold War-era law designed to compel private industry to prioritize government needs during national emergencies. While the DPA was recently utilized during the COVID-19 pandemic for medical supplies, its application to generative AI models marks a significant expansion of executive power into the digital frontier. Furthermore, the threat to label Anthropic a "supply chain risk" is a tactical maneuver usually reserved for foreign adversaries. Such a designation would not only jeopardize Anthropic’s existing government contracts but could also create a chilling effect for private-sector partners and investors who fear regulatory contagion.

“

The Pentagon’s ultimatum is particularly aggressive, leveraging the Defense Production Act (DPA)—a Cold War-era law designed to compel private industry to prioritize government needs during national emergencies.

Anthropic’s refusal is rooted in two specific "red lines": the use of AI for mass domestic surveillance and the deployment of fully autonomous lethal weapons. Amodei’s argument that current AI systems are not yet reliable enough to operate without human oversight reflects a core tenet of "Constitutional AI," the safety-first framework upon which Anthropic was founded. This stance highlights a growing technical and ethical divide: while the military views AI as a critical tool for maintaining a competitive edge against adversaries, safety researchers view the same technology as too unpredictable for high-stakes kinetic environments where errors can lead to catastrophic unintended consequences.