A dispute over who controls AI within classified military systems is at the center of Anthropic’s legal fight with the Pentagon, as the company revealed it cannot manipulate its Claude model once deployed.
In a filing to a US appeals court, Anthropic warned it would not be able to regulate, govern, or intervene in its AI tool at that stage, prompting it to oppose its classified military use.
The assertion is part of its latest 96-page argument against the Pentagon, aimed at contesting the “supply chain risk” designation and the cancellation of a $200 million contract.
The designation marks the first known instance of the US government applying such a label to a domestic company, as it is typically reserved for guarding against foreign threats to national security systems.
Anthropic has already filed separate legal challenges in California federal court and a federal appeals court in Washington, contesting different aspects of the government’s actions.
The company argued the designation is unjustified and amounts to retaliation linked to its stance on limits around autonomous weapons and surveillance. In its lawsuit, it said “no federal statute authorizes the actions taken” by the Pentagon.
The Pentagon, meanwhile, maintains that Anthropic’s technology could pose risks to national security systems and has used the designation to restrict the company’s access to military contracts.
A federal appeals court in Washington has so far declined to block the Pentagon’s actions while the case continues.
