Anthropic spent months negotiating with the Pentagon as a willing defense partner, deploying its AI tool Claude on classified networks for intelligence analysis, cyber operations, and mission planning since June 2024. CEO Dario Amodei drew just two red lines: no mass domestic surveillance of Americans, and no fully autonomous weapons. Current AI, he argued, isn’t reliable enough to make kill decisions without a human in the loop, and warrantless bulk data collection on citizens is incompatible with the democracy the military is tasked with defending.
The Department of War, insisting that it had no plans to violate either condition, nonetheless demanded that those ethical commitments be removed.
When Anthropic refused, the threats escalated, and at the expiration of Secretary of War Pete Hegseth’s Feb. 27 deadline for Anthropic to cave, with the company still standing firm, Hegseth designated Anthropic a “supply chain risk to national security.” Effective immediately, he claimed, no military contractor could do business with the company. Hours earlier, U.S. President Donald Trump had branded Anthropic “Leftwing nut jobs” on Truth Social, ordering all federal agencies to drop its technology. It is bizarre, to say the least, that a technology could be both essential to national security, yet a “supply chain risk” at the same time. The policy used to ban the Chinese firm Huawei is now aimed at an American AI company.
Anthropic’s response was defiant: The designation was “unprecedented,” “legally unsound,” and would be challenged in court. Anthropic proclaimed: “No amount of intimidation or punishment from the Department of War will change our position.”
