Anthropic, DoD Clash in DC Court Over Blacklisting

A pivotal legal battle is set to unfold this week as a federal appeals court in Washington, D.C. prepares to hear arguments in Anthropic’s lawsuit challenging its blacklisting by the Department of Defense. This case represents a critical juncture in the ongoing tension between the Pentagon and one of the nation’s leading artificial intelligence innovators.

On Tuesday morning, a three-judge panel will dedicate 15 minutes to each side: the U.S. Department of Justice, representing the DOD, and Anthropic. The judges – Karen Henderson, Gregory Katsas, and Neomi Rao – will then deliberate and issue a written opinion on the matter. Proceedings are scheduled to commence at 9:30 a.m. ET.

The dispute ignited in March when Anthropic filed suit against Defense Secretary Pete Hegseth and the DOD. The agency had declared the AI startup a “supply chain risk,” a designation typically reserved for foreign adversaries and one that effectively prohibits defense contractors from utilizing Anthropic’s Claude models in their military engagements. This move, experts suggest, raises significant questions about the government’s approach to AI procurement and national security.

This drastic measure followed months of stalled negotiations between Anthropic and the DOD. The core of their disagreement centered on access and control. The Pentagon sought unrestricted access to Anthropic’s AI models for all lawful purposes. Conversely, Anthropic aimed to secure assurances that its technology would not be deployed for fully autonomous weapons systems or for domestic mass surveillance, reflecting a commitment to ethical AI deployment and risk mitigation.

When an agreement could not be reached, Secretary Hegseth moved to blacklist Anthropic, publicly criticizing the company. Anthropic CEO Dario Amodei stated that the company felt compelled to challenge the “supply chain risk” designation in court, citing a lack of viable alternatives.

Despite the DOD’s actions, reports indicate that the department continued to utilize Anthropic’s models for military operations. Furthermore, President Donald Trump publicly suggested that a resolution between the DOD and the AI firm was “possible,” hinting at potential avenues for de-escalation and compromise.

In April, the appeals court denied Anthropic’s urgent request to temporarily suspend the designation, meaning it remains in effect throughout the ongoing litigation. However, recognizing the potential for significant and irreparable harm to the company during the legal process, the judges agreed to expedite the case. This acceleration suggests the court acknowledges the substantial implications of the designation for a company at the forefront of AI development.

In its legal filings, the government argued that Anthropic possesses the capability to “encode limitations” into its AI models, presenting what it described as an “untenable national-security risk.” The defense maintained that Secretary Hegseth determined Anthropic had “undermined the substantial trust required to sustain the relationship,” particularly given the company’s potential to “manipulate its model to enforce its own moral and policy judgments about the military’s appropriate use of the technology.” This stance highlights the Pentagon’s concern over algorithmic bias and the potential for AI to reflect the developers’ values rather than purely operational requirements.

Anthropic, in its own briefs, refutes the notion that it could arbitrarily encode limitations into future models, arguing that this claim provides “no basis” for the “supply chain risk” designation. The company further contends that the actions of Secretary Hegseth and the DOD infringed upon constitutional rights and existing legal procedures. Anthropic’s legal team has urged the court to declare the designation unlawful.

Adding another layer to this complex legal challenge, Anthropic initiated a separate, yet related, lawsuit in federal court in San Francisco. This dual litigation stems from the DOD’s reliance on two distinct designations to justify its “supply chain risk” action, necessitating separate legal proceedings in different jurisdictions. The strategic decision to pursue distinct legal avenues underscores the multifaceted nature of the dispute and the differing legal frameworks applicable to each designation.

In its San Francisco case, Anthropic secured a preliminary injunction. This ruling permits government agencies, other than the DOD, to continue utilizing Anthropic’s AI models while the litigation progresses. The judge’s reasoning in this instance was particularly noteworthy, as it characterized the government’s branding of an American company as a potential adversary for expressing disagreements as an “Orwellian notion” unsupported by governing statutes.

This legal entanglement highlights the burgeoning challenges at the intersection of cutting-edge AI development, national security, and established governmental processes. The outcome of these lawsuits could set significant precedents for how the U.S. government engages with and regulates advanced AI technologies, particularly those developed by domestic companies with potentially differing ethical frameworks.