A U.S. federal judge has ruled that the Pentagon acted unlawfully when it designated Anthropic a national-security “supply-chain risk.” Judge Rita Lin found that the government’s action amounted to retaliation against Anthropic after the company refused to remove safeguards around certain military uses of Claude, particularly mass surveillance of Americans and fully autonomous weapons. The ruling blocks enforcement of the blacklist and represents a major legal victory for Anthropic.
The case is important because the Pentagon did not simply decide to stop buying Claude. It used a powerful supply-chain designation that could have prevented Anthropic from doing business across the federal government and potentially damaged relationships with defense contractors. Judge Lin concluded that the designation was arbitrary and that Anthropic was denied required due-process protections. She also found evidence that the government was trying to make an example of the company because of its criticism of the Pentagon's AI policy.
At the heart of the dispute is a much bigger question: who controls the boundaries of AI used for national security? The Pentagon argues that the military must ultimately control how technology it purchases is deployed, while Anthropic argues that developers can maintain limits around particularly consequential applications. The court's decision does not prevent the Pentagon from choosing another AI provider; instead, it says the government cannot use its contracting power to punish a company for exercising protected rights or expressing disagreement with government policy.
The broader takeaway is that AI safety rules are becoming a constitutional and national-security issue, not merely a corporate policy question. As governments become increasingly dependent on private AI companies for military and intelligence capabilities, conflicts over surveillance, autonomous weapons and human oversight are likely to become more common. The Anthropic ruling establishes an important boundary: governments may decide which AI systems they want to use, but national-security authority does not automatically give them unlimited power to punish companies that impose ethical limits on their technology.
