Pentagon Issues Friday Deadline to Anthropic Over Military AI Access

The confrontation between Defense Secretary Pete Hegseth and Anthropic CEO Dario Amodei represents a watershed moment for the venture-backed artificial intelligence sector. By issuing a Friday deadline for Anthropic to grant unrestricted military access to its Claude models, the Pentagon is signaling that the era of 'voluntary' cooperation between AI labs and the defense establishment is coming to an end. This move places Anthropic, a company founded on the principles of AI safety and constitutional alignment, in a direct collision course with a Department of Defense that is increasingly impatient with ethical restrictions it views as a hindrance to national security.

Anthropic currently stands as the final major holdout among its peers. While competitors like OpenAI, Google, and Elon Musk’s xAI have already integrated their technology into the U.S. military’s new internal network, Anthropic has maintained strict 'red lines.' Amodei’s refusal to budge on two specific areas—fully autonomous military targeting and the domestic surveillance of U.S. citizens—highlights a fundamental philosophical divide. For the Pentagon, these restrictions are seen as 'woke' barriers to efficiency; for Anthropic, they are existential safeguards against the misuse of powerful technology. Amodei’s recent public warnings about AI’s potential to 'stamp out' dissent through mass surveillance suggest that the company views this as more than just a contractual dispute, but a defense of democratic norms.

“

The confrontation between Defense Secretary Pete Hegseth and Anthropic CEO Dario Amodei represents a watershed moment for the venture-backed artificial intelligence sector.

The tools the Pentagon is threatening to use are particularly aggressive. Beyond the cancellation of lucrative government contracts, officials have floated the possibility of designating Anthropic as a 'supply chain risk' or, more significantly, invoking the Defense Production Act (DPA). Traditionally used to compel the manufacturing of physical goods like ventilators or munitions during wartime, the application of the DPA to large language models would set a massive legal precedent. It would essentially allow the government to treat private software code as a strategic resource, potentially overriding the intellectual property rights and ethical guidelines of the developers. This would transform the relationship between the state and the tech industry from one of partnership to one of command and control.