U.S. District Judge Rita F. Lin on Aug. 27 vacated the Pentagon’s supply-chain-risk designation and contractor boycott against Anthropic, ruling that the government’s campaign violated the First Amendment, the Fifth Amendment and federal administrative law.

The decision permanently blocks federal agencies from enforcing the directives issued in February and March. It does not require the Pentagon to buy Anthropic’s technology or prevent agencies from moving to other artificial intelligence vendors through lawful procurement processes.

The court found retaliation, not a documented security risk

In a 59-page order, Lin found that the government acted to punish Anthropic after the AI developer publicly defended restrictions on using its Claude models for mass surveillance of Americans and fully autonomous weapons. The dispute escalated after the Pentagon demanded that its contracts permit “all lawful uses.”

The Pentagon initially argued that Anthropic could alter or shut down a deployed model during a military operation. The court said the government later acknowledged that Anthropic lacks that access once a static model is installed on national-security systems. The record also showed that the technology was no riskier than other “black box” AI models used by the government.

Lin concluded that the designation rested on generalized concerns about whether Anthropic could be trusted, including its public criticism of the Pentagon’s position, rather than evidence that the company posed a sabotage or subversion risk. The administrative record, she wrote, was unusually thin and was assembled only after Defense Secretary Pete Hegseth had directed officials to designate the company a supply-chain risk.

The remedy erases the blacklist and contractor ban

The separate final-relief order vacated the designation and the portion of Hegseth’s directive barring any military contractor, supplier or partner from conducting commercial activity with Anthropic. It also set aside orders from several federal agencies that had begun removing Anthropic products in response to the presidential directive.

The court permanently enjoined the participating agencies from implementing the challenged actions and ordered them to rescind related guidance and instructions. Lin denied the government’s request for a seven-day administrative stay, saying officials had not shown irreparable harm after operating under a preliminary injunction for more than five months.

The judgment closes the case at the district-court level. The Associated Press reported that the government is expected to contest the ruling. A separate, narrower case concerning the Pentagon’s ability to remove Anthropic from defense systems remains pending in Washington.

The ruling draws a line around AI procurement power

For AI companies and federal contractors, the ruling distinguishes ordinary vendor selection from punitive action that reaches unrelated commercial business. The Pentagon remains free to choose rival systems, phase out Claude or negotiate different safeguards, but it must do so within procurement rules and constitutional limits.

The decision also narrows how the government can invoke a national-security label. Lin found that the supply-chain statute applies to defined risks such as sabotage and requires officials to consider less intrusive measures, conduct a qualifying assessment and notify Congress. The record did not show that those steps were completed.

That distinction matters beyond Anthropic. AI suppliers increasingly negotiate government contracts around model controls, surveillance and autonomous systems. The order signals that agencies can evaluate those limits as performance or security terms, but cannot convert a policy dispute into a government-wide punishment without evidence and due process.

Anthropic gains immediate relief, but not a guarantee of restored contracts or continued adoption. Its longer-term position will depend on procurement decisions, any appeal and the unresolved Washington litigation. For the broader market, the ruling puts a judicial check on efforts to use federal purchasing power to isolate a technology vendor from customers outside the disputed contract.