On 1 May 2026, the US Department of Defense signed agreements with seven artificial intelligence companies to integrate their models into its military-grade classified networks (levels 6 and 7). Anthropic, which had until then supplied most of the United States' military AI through Claude, was explicitly left out of these contracts. The exclusion stems from a disagreement over the safeguards governing military use of its AI, in particular autonomous weapons and surveillance.
The most striking element is the label that was applied: the Pentagon designated Anthropic a "supply chain risk," a serious classification normally reserved for foreign entities regarded as adversarial. The company took the matter to a federal court, which placed the designation on emergency hold.
The episode is part of a global race in military AI, driven chiefly by the United States and China and, to a lesser extent, Russia. It highlights an underlying tension between the pace of militarization being sought and the ethical restrictions some technology companies seek to maintain. The signal is a strong one: refusing to lift such restrictions for military use can lead to being treated as an adversarial entity.
How the affair should be read remains disputed. The Pentagon frames the exclusion as a response to an attempt by Anthropic to claim a say over military decisions, while Anthropic maintains it was only defending its safeguards against autonomous weapons and civilian surveillance. Actors also diverge on what the contracts mean — a necessary technological advance for some, insufficiently constrained militarization for others. Several angles remain little covered, including opposition from workers within the sector and the implications of these deals for the laws of war.
