Anthropic seeks fresh court review of Pentagon’s AI blacklist

Anthropic is seeking to reopen a legal fight with the US government over restrictions on military use of its artificial intelligence systems, challenging a court decision that allowed the Pentagon to classify the company as a national security supply chain risk.

The company has petitioned the US Court of Appeals in Washington DC to reconsider its September ruling, which backed the defence department in a dispute over Anthropic’s refusal to allow its technology to be used for autonomous weapons or mass surveillance.

The case has become a significant test of the relationship between the US military and AI developers seeking to impose conditions on how their products are deployed. Anthropic says the Pentagon’s decision has cost it billions of dollars in lost business, although the full financial impact has not been independently established in the information available about the case.

By requesting another hearing, the company is challenging both the government’s use of its legal powers and the court’s assessment of whether the designation was justified.

Anthropic disputes court’s interpretation

The latest challenge follows a divided ruling in September, when two of the three judges on the appeals panel sided with the Pentagon. The majority accepted that Anthropic’s safeguards could hamper military operations and concluded that the department had acted reasonably in responding to the company’s position.

The panel also dismissed Anthropic’s claim that the government had targeted it in retaliation for expressing concerns about AI safety.

Anthropic is now arguing that the majority’s reading of federal law gives the government powers. According to the company’s petition, legislation intended to counter risks from hostile foreign actors should not be interpreted as automatically extending to domestic government suppliers.

The company has also challenged the process used to impose the designation. It argues that the Pentagon was allowed to take action without first providing an opportunity for the company to contest the decision, despite what Anthropic describes as evidence that its position on AI safety had prompted retaliation.

The appeal remains unresolved. A request for a rehearing gives the court an opportunity to reconsider its earlier decision, but does not guarantee that the judges will agree to revisit the case or change their conclusion.

Anthropic has had a separate legal success in California. In August, a federal judge there overturned a related designation imposed under a different law. That judgment addressed a separate legal route and did not remove the need to challenge the decision considered by the Washington appeals court.

AI safety v/s defence priorities

The dispute reflects a growing conflict over who should determine the acceptable uses of advanced AI: the companies that develop the systems or the government agencies that purchase them.

Anthropic has drawn boundaries around the use of its technology in areas it considers particularly sensitive, including autonomous weapons and mass surveillance. The Pentagon’s position, as accepted by the appeals court majority, is that these restrictions may limit its ability to use the technology for military purposes.

The disagreement also reflects competing approaches to AI policy within the United States. CEO Dario Amodei has called for regulation that applies across the country’s leading developers of advanced AI systems. The Trump administration, by contrast, has generally prioritised policies designed to accelerate industry growth.

The court proceedings could help establish how national security authorities may respond when a domestic technology company refuses to meet government demands over the use of its products. They also raise questions about the protections available to contractors that publicly advocate limits on potentially harmful applications of AI.