Anthropic Pushes for Rehearing After Court Upholds Pentagon Blacklist Over AI Safety Dispute

Anthropic Challenges Pentagon Blacklist in Federal Appeals Court

Anthropic filed a petition on Tuesday asking the full U.S. Court of Appeals for the District of Columbia Circuit to rehear its challenge against the Pentagon's designation of the company as a national security supply chain risk1

1

2

2

. The move targets a September 25 ruling where a divided three-judge panel voted 2-1 to uphold the Defense Department's blacklisting of Anthropic's Claude models3

3

. The company's attorneys argue that the D.C. Circuit majority allowed the Pentagon to act without any pre-designation process and ignored straightforward evidence that the designation was retaliation for Anthropic's speech about AI safety4

4

.

The Core Dispute Over AI Safety Commitments

The conflict between Anthropic and the Trump administration stems from a breakdown in negotiations over how the military could deploy the company's AI models. Anthropic had signed a $200 million contract with the Pentagon in July 2025, but talks collapsed that September when the company sought guarantees that Claude would not be used for mass surveillance of Americans or in lethal autonomous warfare systems1

1

. The Defense Department, led by Pete Hegseth, rejected these conditions and subsequently applied the supply chain risk label in March, effectively blocking defense contractors from using Anthropic's technology in their work with the agency1

1

. Anthropic's petition states that the company was branded a risk because it declined to permit its models to be used for mass surveillance or autonomous weapons and publicly explained those AI safety commitments1

1

.

Legal Battle Produces Split Decisions Across Courts

The September ruling by Judge Gregory Katsas, joined by Neomi Rao, concluded that the Pentagon acted reasonably and had legitimate grounds to worry that Anthropic could alter Claude's behavior in ways that would render it unfit for missions the department viewed as essential3

3

. Judge Karen LeCraft Henderson dissented, reading the statute narrowly and arguing that its history concerns hostile states rather than a contractor enforcing its own terms openly2

2

. However, in a parallel case in California, U.S. District Judge Rita Lin ruled in August that the Defense Department had violated Anthropic's First Amendment rights and that the supply chain risk designation amounted to illegal retaliation, ordering the government to rescind all directives issued against the company3

3

.

Financial and Strategic Implications for Anthropic

The Pentagon's blacklisting has cost Anthropic billions of dollars in lost business, according to the company4

4

. While government work represents under 1% of revenue, Anthropic has told IPO investors that the government's view of it could cost material revenue with commercial customers too2

2

. The Pentagon has moved forward regardless, with its chief technology officer stating in September that about 90% of the classified AI workload had been shifted off Anthropic, with expectations to complete the transition by the end of that month2

2

. Before the clash spilled into public view earlier this year, Anthropic served as an early partner across many U.S. agencies, including the Defense Department1

1

.

Broader Context of AI Regulation and Military Use

Anthropic's petition argues that the panel majority misinterpreted and transformed a federal law aimed at hostile foreign actors into one that presumptively reaches every government contractor4

4

5

5

. Historically, such supply chain risk designations have been reserved for firms based in countries the U.S. considers adversaries3

3

. The legal and policy dispute highlights a growing tension between AI companies prioritizing safety commitments and government demands for unrestricted access to AI capabilities for military operations. Anthropic CEO Dario Amodei has called for regulation covering all US frontier AI companies, while the Trump administration has favored a pro-growth approach4

4

. In Europe, this fight could not take the same shape, as the European AI Act does not cover systems used exclusively for military, defence or national security purposes2

2

. European labs like Mistral have drawn their own conclusions, building defence models with firms like Helsing since early last year2

2

.