Astute Beating AI Newsflash: Anthropic has emerged victorious in this round of litigation against the Pentagon. United States Federal Judge Rita Lin ruled that the government's classification of Anthropic as a "supply chain risk," the directive to federal agencies to cease using Claude, and the restriction on defense contractors collaborating with Anthropic were all deemed unlawful. The court concluded that these actions were in retaliation for Anthropic's public criticism of the government's AI policy.

The judgment heavily criticized the evidence presented by the government. The entire rationale behind the set of measures was contained in a mere 4-page memo, which was also dated later than two of the three measures. Previously, the government had expressed concerns that after deploying models, Anthropic could clandestinely alter them through a "backdoor," but trial evidence confirmed that Anthropic did not possess such authority. The court held that the government's remaining justification was simply a lack of "trust" in Anthropic and a desire to assert dominance publicly.

The conflict initially arose from the military usage restrictions imposed on Claude. Anthropic insisted that the model could not be utilized for widespread surveillance of Americans and fully autonomous weapons, while the Pentagon demanded it be available for all lawful purposes. The court also made it clear that the judgment did not compel the military to continue using Claude. The Pentagon can still switch vendors lawfully; however, they can no longer blacklist Anthropic using the "supply chain risk" justification.