A federal appeals court has upheld the Pentagon’s decision to blacklist Anthropic as a national security risk, ruling 2-1 against the AI startup on Friday. The decision preserves restrictions on military contracts following a dispute over safety guardrails for autonomous weapons and domestic surveillance.
The U.S. Court of Appeals for the District of Columbia Circuit delivered a legal setback to Anthropic, siding with the Defense Department in a high-stakes clash over artificial intelligence in military systems. The majority opinion, written by Judge Gregory Katsas, found that the Pentagon possessed valid statutory authority to designate the San Francisco-based company as a supply chain risk, keeping its technology locked out of classified defense networks.
The D.C. Circuit Majority Ruling and Defense Department Authority
Judge Katsas, joined by Judge Neomi Rao, ruled that the Defense Department adequately demonstrated that keeping Claude embedded in its information systems created a risk to national security covered by statute. Judge Karen LeCraft Henderson dissented from the 2-1 decision. Judges Katsas and Rao were appointed by President Donald Trump, while Judge Henderson was appointed by former President George H.W. Bush.
Federal Appeals Court Backs Pentagon’s Anthropic Blacklisting
The court directly addressed the operational concerns raised by Defense Secretary Pete Hegseth, who argued that an AI developer’s internal safety restrictions could jeopardize military readiness. Writing for the majority, Katsas noted the deeply sobering prospect of overly constrained AI models shutting down unexpectedly and thus causing important military operations to fail.
The appellate panel also rejected Anthropic’s arguments that the blacklisting violated its First Amendment rights or amounted to unlawful retaliation. The court concluded that the government acted within its lawful scope based on the company’s refusal to accept contract terms deemed essential by military leaders.
Contract Disputes, Autonomous Weapons, and the $200 Million Deal
The friction between the Pentagon and the AI startup traces back to a $200 million contract dispute over how the Pentagon could deploy Claude on classified systems, after the two first teamed up in July of last year through a deal in which Anthropic agreed to prototype frontier AI capabilities that advance U.S. national security. Tensions boiled over as negotiations stalled over how the technology could be deployed on classified systems.
Federal appeals court upholds Pentagon's Anthropic blacklist
The Pentagon demanded that Anthropic make its models available for any lawful military use without restriction. Anthropic, meanwhile, sought written guarantees that its Claude models would not be utilized in fully autonomous weapons systems or to conduct mass surveillance of the domestic population.
Following the breakdown in talks, Hegseth designated Anthropic a supply chain risk—a classification historically applied to foreign adversaries. That label effectively blocked every Pentagon contractor and supplier from using Anthropic’s technology, which the startup says has cost it billions of dollars in lost business and damaged its reputation ahead of a highly anticipated initial public offering. In February, The Wall Street Journal reported that Claude was used by the U.S. military during its operation to capture Venezuelan president Nicolás Maduro and his wife, Cilia Flores, from a presidential palace in Caracas. Later that month, Anthropic CEO Dario Amodei said that Anthropic would not agree to allow the federal government to use its technology for domestic surveillance or fully autonomous weapons systems, two applications of AI which, according to Amodei, can undermine, rather than defend, democratic values. U.S. Secretary of War Pete Hegseth shot back that Anthropic’s refusal to consent to what his department defined as any lawful use of AI amounted to the company trying to call the shots on U.S. military strategy. In an X post on February 27, he described Amodei’s stance as a master class in arrogance and betrayal as well as a textbook case of how not to do business with the United States Government or the Pentagon.
Contrasting Judicial Outcomes in San Francisco and Washington
Friday’s appellate ruling represents only one front in a fragmented legal battle. Because the Pentagon grounded its blacklisting in two separate legal designations, each required litigation in a different court.
US appeals court upholds Pentagon’s blacklisting of Anthropic
While the D.C. Circuit upheld one designation, a federal judge in San Francisco ruled last month that a parallel DOD designation was unlawful. U.S. District Judge Rita Lin found that the administration had retaliated against Anthropic in violation of the First Amendment over its views on AI safety and ethics, blocking both the government-wide ban on Anthropic and Hegseth’s order barring military contractors from doing any business with the startup.
Anthropic indicated it is weighing its next steps following the D.C. Circuit ruling.
We respectfully disagree with the court's decision,
an Anthropic spokesperson said in a statement. Another federal court has already held the government's parallel designation unlawful. We remain confident in our position and are considering all options, including further review.
The appellate panel temporarily stayed its ruling, preserving Anthropic’s ability to request a panel rehearing, pursue an en banc review by all D.C. Circuit judges, or carry the case to the Supreme Court.
По теме
