Court says Trump may penalize Anthropic after it declined to activate Claude features

A federal appeals court has sided with the Trump administration in its dispute with AI company Anthropic, ruling that the government's blacklisting of the company qualified as a legitimate response to a "supply chain risk" under section 4713 of federal procurement law. The decision is a setback for Anthropic, which had won earlier in district court on First Amendment grounds after refusing to lift restrictions on how its Claude model could be used by the government.

How the Dispute Started

The conflict centers on Anthropic's practice of building usage restrictions directly into Claude. According to the appeals court, Anthropic "encodes restrictions into Claude that prevent the model from performing tasks that Anthropic wishes to prevent." Those restrictions have, on more than one occasion, blocked Claude from completing tasks requested by government users.

The situation came to a head over a contract dispute: the parties disagreed about whether contractual prohibitions barred the use of Claude in an ongoing overseas military operation, leaving the Department uncertain whether the model would perform as intended. Anthropic had also refused to drop restrictions on using its products for lethal autonomous warfare and for mass surveillance of Americans.

Anthropic alleged that the administration then retaliated illegally by blacklisting the company. Judge Rita Lin, a Biden appointee presiding over the case in the Northern District of California, agreed, determining that the blacklisting violated the First Amendment. "The empty invocation of national security is not a blank check to punish and retaliate against government critics," Lin wrote.

Why the Appeals Court Found a "Supply Chain Risk"

Section 4713 is the statutory provision at the heart of the appeal. The court majority concluded that the Secretary's concern about Anthropic disabling Claude from performing lawful actions requested by the Department "qualifies as a 'supply chain risk' within the meaning of section 4713."

Two elements of the statute's language carried particular weight. First, the phrase "any person" shows the definition is not limited to adversaries or foreign entities. Second, the word "deny" applies to a company preventing the US government from using certain features of a product — exactly what Anthropic's built-in restrictions could do.

The majority added that "the Department reasonably feared that Anthropic might manipulate Claude's design to prevent it from performing national-security functions that the Department deems contractually authorized and necessary."

The Dissent: A Statute Meant for Hostile Actors, Not Honest Contractors

Judge Karen Henderson, a George H.W. Bush appointee, dissented. She argued that the majority misread the definition in section 4713: "viewed in their statutory context, the verbs at issue are all directed at deliberately impeding or eavesdropping on the 'function, use, or operation' of a covered article that has entered the federal supply chain."

Hendonson pointed to the law's origins. Congress, she wrote, "enacted the statute in response to calls from the US intelligence community for legislation to meet the threat of '[h]ostile nation state and other bad actors' infiltrating the federal government's information and technology systems through its supply chains." In her view, the definition should not stretch to cover "a contractor's honest and upfront enforcement of restrictions on a covered article's use disfavored by the government."

Where the Case Stands

The ruling reverses the trajectory of the litigation: Anthropic had won at the district court stage, where Judge Lin found the blacklisting unconstitutional, but the appeals court has now accepted the government's national-security framing under section 4713. The core tension remains unresolved — whether an AI vendor can enforce its own usage limits on a government customer without triggering supply-chain enforcement, and where First Amendment protections end when national security is invoked. Watch for a possible petition for further review and for how other AI providers structure government contracts in light of this ruling.