The US Court of Appeals for the District of Columbia Circuit ruled on September 25, 2026, that the Trump administration's blacklisting of Anthropic is lawful under the Federal Acquisition Supply Chain Security Act of 2018. The decision supports the Department of War's authority to exclude Anthropic from the federal supply chain because of the company's refusal to provide "all lawful use" access to its Claude AI models.
US Appeals Court Rules Trump Administration Can Blacklist Anthropic for Usage Policy Restrictions
PLUS ULTRA by Amenoyomi
The dispute stems from Anthropic's decision to maintain contractual prohibitions on certain uses of Claude, specifically regarding lethal autonomous warfare and mass domestic surveillance. Anthropic argued that these restrictions are core to its mission of ensuring AI safety and preventing misuse. However, the court found that the Department of War's decision to restrict the company qualifies as a "supply chain risk" under section 4713 of the Act.
The majority opinion noted that Anthropic's model training effectively allows the company to enforce these restrictions, which creates uncertainty for the Department regarding whether Claude will perform as intended during critical military operations. The court held that the government's concern about Claude potentially being unable to perform required tasks due to these safeguards is a reasonable basis for the designation.
The ruling was split 2-1, with Judge Karen Henderson dissenting. Judge Henderson argued that the term "manipulate" in the statute should be interpreted as requiring a malicious or subversive intent, similar to terms like "sabotage" or "maliciously introduce." She contended that the administration's action was an attempt to punish the company for its policy stances rather than addressing actual security threats.
Anthropic has indicated it may appeal the decision to the Supreme Court or seek an en banc review by the full appeals court.
PLUS ULTRAby Amenoyomi
The legal basis for this decision rests on the specific definition of "supply chain risk" under 41 U.S.C. § 4713. Unlike some other procurement laws, this statute does not require the actor to be a foreign adversary or to possess malicious intent. Instead, it defines a risk as the possibility that "any person" may manipulate the design or operation of a product so as to "deny" its function.
In this case, the court interpreted Anthropic's safety guardrails—which the company implements to prevent the AI from being used for lethal autonomous warfare or mass surveillance—as a form of "denial." From the government's perspective, these restrictions created an unacceptable uncertainty: the risk that the AI might unexpectedly shut down or refuse to perform a task during a critical military operation, despite that task being lawful and contractually authorized.
This interpretation explains why the appeals court reached a different conclusion than the lower district court. The district court had previously ruled the blacklisting illegal by applying 10 U.S.C. § 3252, a narrower law that limits supply chain risks to "malicious" actions by "adversaries." However, the appeals court held that because the administration invoked the broader authority of § 4713, the absence of bad motive was irrelevant to whether the company's restrictions constituted a legal risk to national security.
Sources
- Court rules Trump can blacklist Anthropic for refusing to enable Claude features （Ars Technica AI, 2026-09-25）
- US Court of Appeals ruling