Pentagon and Anthropic Clash Over AI Access

The Korean War-era Defense Production Act (DPA) grants the president broad authority to direct private industry for national defense, a power previously used to compel production of everything from ventilators during the COVID-19 pandemic to baby formula. Invoking it to force a software company to remove ethical guardrails from its AI model would be an unprecedented expansion of the act's authority, moving from compelling the production of goods to dictating the fundamental design of a private company's intellectual property. At the heart of the dispute are Anthropic's "red lines": a refusal to allow its technology to be used for fully autonomous weapons systems or for mass surveillance of U.S. citizens. These restrictions are part of a contract valued at up to $200 million that Anthropic, along with Google, OpenAI, and xAI, was awarded to provide AI capabilities to the Department of Defense. While the Pentagon argues it only requires the AI for lawful activities, Anthropic CEO Dario Amodei has held firm on these specific limitations. This U.S. strategy presents a stark contrast to the European Union's approach. The recently passed EU AI Act explicitly excludes AI systems developed or used exclusively for military and defense purposes from its scope, creating a clear boundary between civilian and military regulation. This divergence highlights a growing fragmentation in global AI governance, forcing multinational tech companies to navigate conflicting legal and ethical frameworks. The Pentagon's effort to integrate these models is centered on a new platform called GenAI.mil, designed to roll out generative AI tools across the department. Anthropic was the first and, for a time, the only AI provider approved for use on classified military networks, making its potential designation as a "supply chain risk" a significant disruption to the Pentagon's own AI adoption roadmap. While not directly about silicon, this conflict echoes the U.S. government's broader strategy of intervening in the technology supply chain for national security, as seen with the CHIPS Act. That legislation uses billions in federal subsidies to re-shore semiconductor manufacturing, linking private industry's production capabilities directly to defense needs and demonstrating a willingness to use federal power to secure critical tech infrastructure. Should the Pentagon proceed, it could force Anthropic to deliver its Claude AI model without the current contractual restrictions, a move legal experts say is without precedent under the DPA's history. The government's leverage is substantial, including not just the DPA threat but also the possibility of canceling the lucrative contract and officially designating Anthropic a "supply chain risk," which would impact its ability to work with other government contractors.