1. What did Anthropic ask the appeals court to do?
Anthropic, the developer of the Claude family of AI models, asked the full U.S. Court of Appeals for the D.C. Circuit on October 13, 2026, to rehear a case it lost in September. The ruling at issue allows the Defense Department to keep treating a domestic AI developer as a national security risk, a status that bars Pentagon personnel and contractors from using Claude. The request revives the Anthropic Pentagon lawsuit at the second-highest level of the federal court system before any move toward the Supreme Court.
The filing is a petition for rehearing en banc. In ordinary appeals, a panel of three judges hears the case and issues the decision. An en banc petition asks all of the court's active judges to review that decision instead, usually on the argument that the panel erred on an important question of law. Courts grant such petitions only rarely, so the request is an uphill effort by design, but it is a required stop for many litigants who intend to continue their fight.
Anthropic is asking the court to set aside the panel's outcome and look at the dispute again with every active judge participating. The company's central claim is that the three-judge panel misapplied the governing law and did not properly address the due process and First Amendment questions raised by the Pentagon's conduct toward the company.
2. How did the three-judge panel rule?
The panel issued its decision on September 25, 2026, and it favored the Pentagon. At the center of the case is a designation under the Federal Acquisition Supply Chain Security Act, known as FASCSA, a law that lets federal agencies exclude vendors whose products are judged to pose a risk to government supply chains. The Pentagon labeled Anthropic a supply chain risk under that statute, which in practice shuts Claude out of defense work.
The government's stated worry concerns the terms under which Anthropic licenses its technology. Those terms prohibit the use of its systems for lethal autonomous warfare and for domestic surveillance. Defense officials have argued that restrictions of this kind could limit what the military is able to do with a tool it depends on, and that the possibility alone is a legitimate security concern.
The majority of the panel agreed. It found what it described as ample support for the designation and concluded that the limits Anthropic places on how its models can be used could constrain military operations. Judge Henderson dissented from that result. The source material does not set out the reasoning of her dissent, but her disagreement is the foundation of the argument that the question deserves a second look by the full court.
3. What is the legal argument for rehearing?
Anthropic's petition rests on two lines of criticism. First, it says the panel got the law wrong in how it applied the supply chain statute to a domestic AI developer. Second, it contends that the panel did not meaningfully engage with the constitutional claims that run through the case.
The First Amendment claim holds that the government acted against Anthropic because of positions the company has taken in public on AI safety, rather than because of any real defect in its products. The due process claim concerns the manner in which the designation was imposed and whether the company received a fair opportunity to respond. Both arguments say the Pentagon's action was punishment for speech and for contract terms, not a neutral security judgment.
The distinction between the two layers of review is worth keeping in mind. A panel decision binds the circuit unless it is overturned, which is why litigants who lose before three judges often turn to the full court before anything else. The rules for these petitions are demanding because the full bench is expected to spend its time on questions of exceptional importance, not on routine disagreements about how a statute was read.
An en banc petition does not retry the facts. It has to persuade the full bench that the panel's decision conflicts with precedent or decides an exceptionally important legal question wrongly. By pointing to constitutional issues it says were left unanswered, Anthropic is trying to fit its request into that narrow category.
4. How did a contract dispute become a security designation?
The conflict began in early 2026, when Anthropic and the Pentagon negotiated over how the military could use the company's AI. The talks centered on the same restrictions on autonomous weapons and surveillance that now sit at the heart of the lawsuit, and they stalled without an agreement.
After the negotiations broke down, the Trump administration escalated the matter by labeling Anthropic a national security risk. That step is regarded as significant and unprecedented for a domestic AI company, because supply chain risk tools of this type are more commonly associated with vendors tied to foreign adversaries. The label turned a commercial disagreement over contract language into a formal exclusion from the defense market.
The sequence matters to the legal fight. If the designation grew directly out of a stalled negotiation over a company's usage policy, Anthropic can argue that the government responded to a refusal rather than to a demonstrated threat. The government, for its part, can argue that a supplier's insistence on limiting military uses is itself the risk it is entitled to guard against.
5. Why do two courts disagree on the same conduct?
The D.C. Circuit decision is not the only judicial word on the government's treatment of Anthropic. In August 2026, Judge Rita Lin of a federal district court in California ruled that related Pentagon actions were unlawful retaliation against the company for its public positions on AI safety. Her decision, issued on August 27, described the government's conduct as unconstitutional.
The two courts therefore examined connected conduct and reached opposite conclusions. The California court treated the actions as punishment for protected speech. The panel in Washington treated the matter as one calling for deference to national security judgments. Anthropic won in California and lost in Washington.
That divergence is the practical engine of the rehearing petition. Anthropic is relying on the gap between the rulings to argue that the issue is serious enough for the full D.C. Circuit. The core of Judge Lin's reasoning has not been challenged so far, which leaves the company with a favorable decision on the retaliation question even as the designation itself stands in the capital.
6. What does the designation mean for Anthropic today?
For now, the most immediate consequence is access. While the designation remains in force, the Defense Department and its contractors cannot use Claude. Because many technology firms hold federal contracts, the ban's practical reach can extend beyond the Pentagon's own staff into the companies that serve it, which have to decide whether to keep Claude out of any work connected to defense.
The effect on a company's business is hard to measure from the public record, and the source material does not put a dollar figure on it. What the record does show is the structure of the dispute: Anthropic's usage restrictions are the reason for the exclusion, and the exclusion is the reason for the litigation. A lasting resolution would likely require either a court decision that changes the status of the designation or a negotiated settlement over the contract terms that began the confrontation.
There is also a question of precedent. Because the designation tool was designed for supply chain security, a ruling that lets it be applied to a U.S. developer over a licensing disagreement would stand as guidance for future disputes between agencies and technology suppliers. A ruling that narrows its use would do the opposite. Either way, the reasoning the full court chooses to adopt, or declines to review, will be read closely by lawyers advising both government buyers and AI vendors.
The case also carries weight for the wider AI industry. Other model developers write usage policies of their own, and the outcome will help show how much room a company has to set limits on military use without risking exclusion from government work.
7. What happens next in the case?
The immediate question is procedural: the D.C. Circuit must decide whether to grant rehearing. If a majority of active judges votes to do so, the panel's September ruling could be reconsidered from the beginning, and the full court would have to weigh national security deference against the constitutional protections Anthropic invokes.
If the court declines, the panel's ruling stays in place within that circuit. The company would then face the choice of whether to seek review at the Supreme Court, and two federal courts would continue to hold sharply different views of the same government campaign. A conflict of that kind is often a reason for the Supreme Court to take an interest, though it is under no obligation to do so.
Timing is another open point. Rehearing petitions can be resolved within weeks, but nothing in the available record fixes a schedule, and the court may call for a response from the government before it acts. Even a denial would not end the matter, since the California ruling remains on the books and the underlying dispute over contract terms would still be unresolved.
No date has been set for a decision on the petition. Observers will be watching for the court's order on rehearing, for any response the government files, and for further proceedings in California that could bear on the designation. Until one of those developments arrives, the Pentagon's ban on Claude remains the operative reality.