Anthropic Seeks En Banc Review: Washington Circuit Previously Upheld Blacklist 2-1

Anthropic has asked the full panel of the Washington Court of Appeals to re‑examine a decision that upheld the blacklist, a decision that was previously affirmed by a 2‑1 vote.²³ The blacklist designates Anthropic’s Claude model as a national‑security supply‑chain risk, effectively prohibiting defense contractors from using the company’s technology.¹³ Thus, the petition disputes not only whether the court should maintain the original decision, but also whether the government may apply a risk label to Anthropic and how the court interprets its safety commitments.¹ Supported reasoning holds that one goal of the petition is to challenge the court’s reading of Anthropic’s technology‑safety pledge and the scope of the risk‑label application.¹

The episode’s evolution includes the government’s restriction, commercial cooperation between the parties, and a subsequent breakdown in deployment negotiations.³ CNBC Technology reported that the Pentagon tagged Anthropic as a supply‑chain risk in March 2023, preventing defense contractors from using its technology.³ Later, Anthropic signed a $200 million contract with the Pentagon in July 2025.³ In September of the same year, negotiations over Claude’s deployment fell apart, and the relationship soured thereafter.³

Another court‑decision report dated the ruling to September and said the court found Claude to be a national‑security supply‑chain risk, blocking defense contractors from using Anthropic’s technology.¹ That material does not specify the year; the March 2023 date cited by CNBC Technology is when the Pentagon applied the label.¹³ According to CNBC Technology, the Washington Circuit upheld the blacklist in September 2023 by a 2‑1 vote and rejected multiple legal challenges Anthropic raised against the ban.³ These records refer separately to the label, the ruling, and the negotiations and should not be collapsed into a single event or point in time.¹³

The measure directly affects defense contractors that wish to use Anthropic’s technology, while Anthropic and its Claude model are the entities labeled as risky.¹³ The reporting describes the restriction as contractors being unable to use the company’s technology, without providing numbers of affected contractors or a specific list.¹³ The court’s vote ratio of 2‑1 indicates the difference between the majority and dissenting opinions when the decision was upheld.³ The $200 million figure refers to the Anthropic‑Pentagon contract amount, not to the scope of the risk label or a count of contractors.³

Anthropic argues that the court misinterpreted federal law by extending provisions originally aimed at hostile foreign entities to all government contractors.¹ The company also claims the ban is arbitrary, lacks authorization, and is unconstitutional; the appellate court rejected those arguments in its September 2023 decision.³ The current en banc petition therefore seeks to have more judges re‑consider the prior appellate ruling.²³ Existing material does not indicate whether the full panel will accept the petition nor does it provide a review outcome.²

Anthropic further alleges that the Department of Defense acted before completing its identification process and says evidence shows the move was retaliation for the company’s AI‑safety statements.² These are the company’s accusations, not facts that the court has already confirmed.² The majority opinion’s presented Department‑of‑Defense concern is that Anthropic could modify Claude so the model would be unable to perform tasks the Defense Department deems necessary for national security.² Both sides differ on whether procedures were sufficient, how risk should be assessed, and how safety positions influence government decisions.²

The majority view holds that the Pentagon’s action was reasonable, citing Anthropic’s refusal to allow its technology to be used for autonomous weapons or mass surveillance.¹ Judge Katsas, writing for the majority, noted the Defense Department’s worry that Claude might be altered, rendering it unable to carry out tasks the department considers essential for national security.² This passage explains the government’s concerns about the model’s use and availability, but the material does not state that Anthropic has actually made such modifications.² The rationale in the majority opinion differs from Anthropic’s claim of retaliation over its safety stance; the latter remains the company’s assertion.¹²

Before the relationship soured, the contract provided another collaborative context: Anthropic and the Pentagon signed a $200 million agreement in July 2025.³ Later that September, negotiations over Claude’s deployment collapsed, and reporting linked the deteriorated relationship to that failed negotiation.³ These details show that the contract cooperation and deployment dispute co‑existed, but they do not explain whether the contract was terminated because of the negotiation breakdown.³ Therefore, one cannot infer from the contract amount alone that the contractor ban has been lifted or that the Pentagon has resumed using Claude.¹³

The institutional backdrop of this dispute is the government’s supply‑chain risk labeling and its impact on contractors’ use of technology, as well as the court’s review of the legal basis and procedure for that measure.¹²³ Anthropic’s en banc petition seeks to re‑argue the scope of the statutory language and to question how the court handled its safety commitments and the risk label.¹² The prior appellate ruling upheld the blacklist by a 2‑1 vote, indicating the decision was not unanimous among the judges, though the material does not provide the dissenting opinion’s specific reasoning.³ Current material presents the petition, the existing ruling, and the parties’ points of contention, but does not state whether the review has been granted or what outcomes may follow.²³