DC Circuit Upholds Pentagon Anthropic AI Supply Chain Risk

• A three-judge panel of the D.C. Circuit Court of Appeals upheld the Pentagon's designation of Anthropic's AI products as a "supply chain risk."
• This ruling permits the Department of Defense to ban Anthropic AI use by its personnel and private contractors working on defense projects.
• The majority opinion cited Anthropic's embedded restrictions in its Claude AI as a basis for the national security risk determination under FASCSA.
• A dissenting judge argued that the Federal Acquisition Supply Chain Security Act of 2018 (FASCSA) was not designed for a U.S. company building safety guardrails.
• The decision does not affect a separate California lawsuit where a judge ruled against a broader ban on Anthropic from all federal contracts, and further appeals to the Supreme Court or an en banc D.C. Circuit panel are possible.

Lawyers and compliance officers advising federal defense contractors must urgently assess their use of Anthropic AI, as the DC Circuit's upholding of the Pentagon's 'supply chain risk' designation under FASCSA restricts its use for DoD-related work and sets a precedent for AI product evaluations.

A three-judge panel of the U.S. Court of Appeals for the District of Columbia Circuit has affirmed the Pentagon's controversial classification of Anthropic's artificial intelligence products as a "supply chain risk" to national security. This pivotal decision grants the Department of Defense (DoD) the authority to prohibit the use of Anthropic AI not only by its own personnel but also by private sector employees engaged in defense contracts. The ruling marks a significant development in the ongoing scrutiny of AI technologies within government operations, particularly concerning the Pentagon Anthropic AI supply chain risk.

Following the court's announcement, Pentagon Chief Technology Officer Emil Michael publicly celebrated the outcome, asserting that Anthropic now poses a "Supply Chain Risk to the defense industrial base." He further commented that the ruling would reassure warfighters, ensuring that no private company could impose its views within the military chain of command. Conversely, Anthropic expressed its disagreement with the court's findings, indicating that it is actively evaluating all available avenues, including the possibility of further judicial review, to challenge the decision.

The majority opinion, penned by Judge Gregory Katsas and joined by Judge Naomi Rao, both appointees of former President Trump, concluded that the Department of Defense had substantial grounds to determine that integrating Anthropic's Claude AI into its information systems, whether by the DoD itself or its contractors, constituted a statutorily defined national security risk. The judges highlighted Anthropic's own admission that it incorporates restrictions into Claude, preventing the model from executing tasks the company deems undesirable. These embedded limitations have, on multiple occasions, reportedly hindered government users from performing requested functions.

In a dissenting opinion, Judge Karen Henderson, an appointee of former President George H.W. Bush, argued that the Federal Acquisition Supply Chain Security Act of 2018 (FASCSA) was not intended to address the scenario presented by Anthropic. She contended that FASCSA was explicitly enacted to safeguard against sabotage by malicious foreign entities, not to restrict a U.S. company that openly builds safety and ethical guardrails into its products. This disagreement underscores the complex interpretation of FASCSA AI restrictions when applied to advanced technological products developed by domestic firms.

While the DC Circuit's ruling establishes an Anthropic DoD ban, it is crucial to note that this decision does not impact a separate, parallel lawsuit in the Northern District of California, where a significant ruling has already been made. In that case, Judge Rita Lin recently ruled against the Trump administration's broader attempt to prohibit Anthropic from securing any federal contracts, distinguishing it from the DoD-specific ban upheld in Washington. This divergence highlights the varied legal interpretations and outcomes concerning AI national security risk across different jurisdictions.

The unique procedural path of this case, stemming from the specific law invoked, meant it proceeded directly to the DC Circuit, an appellate court. Legal experts, including Charlie Bullock of the Institute for Law & AI, explain that subsequent appeals are limited to the Supreme Court, which exercises discretion in accepting cases, or an en banc review by the full DC Circuit panel. Sean Timmons, a former JAG Corps officer, suggests that if an en banc review is granted, the outcome could be highly unpredictable due to the diverse political appointments—spanning Trump, Biden, Obama, and Bush administrations—among the full court, contrasting with the two Trump appointees on the initial three-judge panel. This uncertainty means federal contractor AI compliance with the new restrictions remains a dynamic area, with the potential for further legal challenges and shifts in policy.

Lawyers and compliance officers advising federal defense contractors must urgently assess their use of Anthropic AI, as the DC Circuit's upholding of the Pentagon's 'supply chain risk' designation under FASCSA restricts its use for DoD-related work and sets a precedent for AI product evaluations.

Source: Original reporting via Breaking Defense