Judge questions evidence behind U.S. ‘supply-chain risk’ label for Anthropic

Lead

A legal fight over whether Anthropic can be treated as a “supply-chain risk” by the U.S. government has taken a notable turn. According to TechCrunch, a federal judge said during a Thursday hearing that the Trump administration has not yet presented enough evidence to justify the designation or the related ban on federal use of Anthropic’s AI technology.

The case is more than a procurement dispute. It cuts to a central question in the AI era: how much control should a model provider have over the downstream use of its systems, especially when the customer is the military?

Key points

• The evidence is under scrutiny. The judge indicated that the government has not sufficiently demonstrated that Anthropic poses a real supply-chain risk.
• The conflict grew out of DOD contract talks. Anthropic said it did not want its AI used for mass surveillance of Americans or for targeting and firing decisions involving lethal weapons, arguing that the technology was not ready for such use.
• The Pentagon pushed back. The Department of Defense argued that a private company should not dictate how the military uses technology and said it would use the tools in lawful ways.
• Retaliation concerns are central. The government also argued that Anthropic’s public criticism of the DOD helped justify the ban. The judge called that logic deeply troubling because it could set a precedent for retaliating against contractors that disagree with an administration.
• The “kill switch” claim remains unsupported. The DOD claimed Anthropic might disable or alter models during wartime operations, but the judge said she saw no proof that Anthropic could modify a delivered model or trigger such a switch.

Why it matters

The immediate question is whether Anthropic can continue to be blocked from federal use under a risk label. The broader issue is whether the government can use national-security and supply-chain language to exclude AI vendors without a clear evidentiary basis.

For the AI industry, the dispute highlights an unresolved tension. Companies increasingly publish policies limiting uses of their models, especially in areas such as surveillance, weapons, and high-stakes decision-making. But when those companies seek government contracts, those limits may collide with agencies that want operational flexibility.

For government buyers, reliability and control are legitimate concerns. Military users need to know whether a vendor’s systems will remain available and predictable in sensitive situations. But the court’s focus is on proof, not speculation. If the government claims that a company can alter or disable delivered AI systems, it must show evidence that such a capability exists.

The hearing is part of one of two lawsuits Anthropic filed against the Department of Defense in March. The judge had already temporarily blocked the ban and is now considering whether to make that order permanent. Whatever the outcome, the case could become an important marker for how AI safety commitments, federal procurement rules, contractor speech, and military technology adoption intersect.

Source: TechCrunch AI