Judge says Trump admin still lacks evidence for Anthropic ‘supply chain risk’ label

A federal judge has questioned whether the Trump administration has shown enough to support its decision to designate Anthropic a supply-chain threat, a finding that puts renewed pressure on the government’s effort to bar federal use of the artificial-intelligence company’s tools. At a Thursday hearing, the judge indicated that the evidence presented so far did not adequately establish the basis for that designation. The comments cast doubt on the government’s ability to maintain its proposed exclusion on the record currently before the court.

The assessment does not by itself settle the underlying dispute, but it signals skepticism about a central premise of the administration’s position. A supply-chain-risk determination can carry major practical consequences when the federal government is considering whether it may procure or deploy a company’s technology. In this case, the contested designation is tied to an attempted prohibition on the government’s use of Anthropic’s AI systems. The immediate issue is not a general judgment on AI policy, but whether this company and these tools meet the government’s asserted risk standard on the evidence offered.

Bloomberg and Axios were among the first news organizations to report the development. The hearing brought a legal dimension to a conflict that had previously centered on the boundary between a technology provider’s safety limits and the Defense Department’s authority to decide how lawfully obtained tools may be used. The judge’s comments focus attention on whether the government can substantiate the particular rationale it chose for restricting Anthropic. They also distinguish the administration’s asserted supply-chain case from the policy disagreement that led the parties into conflict.

The confrontation grew out of contract talks between Anthropic and the Department of Defense that failed to reach agreement. Anthropic sought limits on certain military and domestic uses of its technology. The company objected to the prospect of its models being used for broad surveillance of people in the United States, and it also resisted use in decisions about selecting targets or launching lethal force. Those points became central because they defined the uses that Anthropic was unwilling to endorse as negotiations continued.

Anthropic’s position was not simply a rejection of government work. Its stated concern was that the technology was not sufficiently mature for the most consequential applications under discussion. By drawing lines around surveillance and weapon-related decision-making, the company argued that deployment should reflect the capabilities and limits of the systems rather than assume that an AI product is ready for every possible government task. The dispute therefore concerns both permitted uses and the company’s assessment of what the technology can responsibly support.

The Pentagon took a different view of the impasse. Defense officials argued that a private vendor should not be able to dictate the purposes for which the military employs technology. The department also maintained that it would use the tools in ways it considers lawful. That disagreement places the two sides at odds not only over a contract, but over who sets the operating boundaries when commercially developed AI enters national-security work. Anthropic’s proposed restrictions and the Pentagon’s claim of operational discretion represent competing answers to that question.

For the administration, the case tests whether a national-security label can withstand scrutiny when it is used to exclude a technology supplier from federal adoption. For Anthropic, the dispute raises the cost of attaching conditions to a sensitive government relationship. The judge’s remarks do not establish that the company’s preferred restrictions must be accepted, but they do suggest that the government will need a clearer factual foundation if it seeks to preserve the supply-chain designation. That is a narrower point than deciding whether the Pentagon’s proposed uses are appropriate, yet it could shape what happens to the restriction.

The next stage will turn on how the administration responds to the court’s concern and whether it can supply additional support for its position. It will also show whether the parties can move beyond the failed negotiations or remain locked in a broader argument over responsible AI use. Any further legal action is likely to keep the focus on the government’s evidence as well as on the contested boundaries Anthropic sought in its agreement. Until then, the federal ban on Anthropic technology faces a more uncertain legal outlook, while the underlying questions about surveillance, lethal-force decisions and military control remain unresolved.

Source: TechCrunch