The D.C. Circuit Court of Appeals is set to hear arguments this week concerning Pete Hegseth’s designation of AI developer Anthropic as a “supply-chain risk” for the Department of Defense. The designation has sparked controversy due to alleged First Amendment violations, questions about executive authority, and concerns over national security implications. Anthropic argues that Hegseth’s actions were retaliatory, legally unsupported, and factually flawed.
Background of the Case
In January, Hegseth publicly labeled Anthropic a “supply-chain risk” after the company expressed concerns that its AI model, Claude, “cannot safely or reliably be used for autonomous lethal warfare and mass surveillance of Americans.” Hegseth also issued a directive barring other military contractors from working with Anthropic, which had no legal effect, and criticized CEO Dario Amodei for “sanctimonious rhetoric.” Anthropic contends these actions constitute “textbook retaliation” against its protected speech and exceeded the Secretary’s legal authority under federal law.
Alleged First Amendment Violations
Federal judges have previously blocked some of Hegseth’s actions over First Amendment concerns. In March, a California judge ruled that Hegseth engaged in “classic illegal First Amendment retaliation.” Anthropic’s petition to the D.C. Circuit asks the court to vacate Hegseth’s supply-chain designation, arguing it is pretextual, retaliatory, and unsupported by law. The company claims that the designation was based on false premises and shifting justifications that undermine its credibility.
Shifting Justifications and “Retreat” Allegations
Anthropic asserts that Hegseth retreated from key factual claims used to justify the designation. Initially, he argued that Anthropic had an “operational veto” over Claude after deployment in classified systems, a claim Anthropic says is demonstrably false. Hegseth later shifted his justification, citing potential future manipulations of Claude prior to deployment as a threat. The petition calls this “shifting justification” evidence that the designation is legally deficient and retaliatory.
Upcoming D.C. Circuit Panel
The oral argument will be heard by a three-judge panel including U.S. Circuit Judges Karen LeCraft Henderson, Neomi Rao, and Gregory Katsas. Henderson has shown some receptiveness to government arguments in prior proceedings, while Rao and Katsas are considered reliably deferential to executive branch national security determinations. Hegseth’s lawyers have emphasized the importance of judicial deference to the Secretary’s “considered national-security judgment” in assessing AI-related risks.
Government Defense Arguments
The Department of Justice argues that Hegseth’s designation focuses on national security concerns, not protected speech. They assert that any errors in timing or sequencing are “harmless” and that predictive judgments regarding AI models in defense applications fall within the apex of executive deference. DOJ attorneys maintain that the Secretary’s determination is grounded in legitimate security concerns over Anthropic’s conduct and the use of Claude in military systems.
Anthropic’s Counterarguments
Anthropic argues that the designation is unlawful, pointing to contradictory actions such as allowing Claude to remain in military use for up to six months despite declaring the company a supply-chain risk. The company cites former Judge Brett Kavanaugh’s precedent to argue that Hegseth’s actions are clearly deficient and must be vacated. The petition emphasizes that the Secretary’s shifting justifications and factual errors render the designation legally untenable.
Broader Implications
The case raises critical issues at the intersection of technology, law, and national security. It tests the limits of executive authority over contractors developing emerging AI technologies, the scope of First Amendment protections for private companies, and judicial deference to national security determinations. Observers note that the ruling could set a precedent for how AI contractors are regulated when their products intersect with defense applications.
