Judge Questions Pentagon’s Blacklisting of AI Firm Anthropic
In a significant legal showdown, a federal judge in San Francisco expressed skepticism regarding the Department of Defense’s (DoD) decision to blacklist the artificial intelligence company, Anthropic. During a hearing that lasted nearly 90 minutes, U.S. District Judge Rita F. Lin scrutinized the government’s actions, suggesting they may conflict with the principles of the First Amendment. This case has garnered attention not only for its implications for Anthropic but also for broader issues of national security and free speech.
The legal battle began after the Pentagon designated Anthropic as a "supply-chain risk" in March, which the company perceives as a threat to its revenue—estimated in the billions of dollars. Both parties convened in the Northern District of California to request Judge Lin to expedite the proceedings and render a decisive ruling. The stakes are high, with the Pentagon arguing its need for autonomy in national security decisions, while Anthropic emphasizes the potentially devastating economic repercussions of the blacklisting.
In her opening remarks, Judge Lin articulated her concerns succinctly, stating, "I don’t see how the additional evidence from the government would help; in some ways, it’s almost worse for the government." Her statement underscored her view that the Pentagon’s actions could be interpreted as retaliatory against a contractor, which she described as troubling and inconsistent with First Amendment rights.
Despite the judge’s probing comments, federal attorneys sought to divert attention from First Amendment issues—a contentious subject for the government, particularly following a preliminary injunction she issued in March. In this ruling, Lin characterized the government’s behavior as "Orwellian," suggesting that the blacklisting may be a form of retaliation against Anthropic for its public stance on responsible AI use.
James Harlow, a senior trial attorney with the Department of Justice, attempted to frame the situation differently. He contended that the government was merely acting within its capacity as a contractor and focused on internal IT systems, asserting that this was more about communication with contractors than any form of regulatory action.
The lawsuit was filed after Anthropic sought to place limits on how its AI models, particularly the Claude models, could be utilized, especially in relation to the newly launched GenAI.mil platform. Anthropic had already secured a $200 million contract over two years in July 2025 but had fundamental concerns about ensuring that its models would not be employed for mass surveillance or lethal autonomous systems.
Michael Mongan, representing Anthropic, argued that the DoD’s justification—citing national security concerns—served as a cover for enforcing a policy the company disagrees with. The Pentagon insists on its right to utilize AI models for "all lawful uses," which Anthropic sees as a threat to its ethical standards.
A significant point of contention arose during the proceedings regarding the nature of AI technology itself. The government contended that AI differs fundamentally from traditional software, which can be regularly updated through patches. Harlow emphasized that AI models often function as "black boxes," where new features or behaviors cannot be easily discerned by users. This necessity for trust poses unique challenges for agencies contemplating the use of AI technology.
In contrast, Mongan countered that the complexities associated with AI apply to all providers and not solely to Anthropic. This encapsulates a wider concern within the industry regarding the ethical deployment of advanced AI technologies.
Judge Lin also probed whether any government agencies had terminated contracts with Anthropic or expanded their use of Claude since her preliminary ruling. Federal attorneys gave evasive answers regarding the status of the company’s models, revealing that while some departments had pulled back on Claude, it would be dangerous to itemize which advanced models the government employs due to security concerns.
As the hearing progressed, it became clear that Defense Secretary Pete Hegseth’s public statements regarding the blacklisting had created confusion. Hegseth had suggested that the ban would extend to all contractors and partners of the U.S. military, but government lawyers maintained that his assertions were not legally binding and that there had not been a secondary boycott against Anthropic.
In light of the complexities of this case, Judge Lin indicated that she would issue a written ruling, leaving both parties awaiting the next chapter in this unfolding legal drama. This case carries significant implications not only for Anthropic but also for the relationship between national security and the burgeoning AI sector, as industries around the world grapple with ethical and operational challenges in the age of artificial intelligence. The outcome of this ruling could set a precedent for future interactions between the U.S. government and technology firms in a rapidly evolving landscape.