Trump administration illegally retaliated against Anthropic, judge rules
Developing story first seen 2 hours ago
A US judge has ruled that the Pentagon illegally retaliated against AI company Anthropic by classifying it as a supply-chain risk after it restricted military uses of its technology. The decision strengthens the finding that national-security powers cannot be used to punish a company for opposing particular government uses of AI or for expressing those views publicly.
Judge Rita Lin said the designation was “illegal and baseless” and that invoking national security was not a licence to retaliate against critics. The Pentagon made the designation in February, reportedly the first public use of the measure against a US company, after Anthropic refused to allow its Claude models to be used for domestic surveillance or autonomous weapons; the firm said the action harmed its business and free-speech rights.
- Judge rules Pentagon illegally retaliated against Anthropic
- Anthropic opposed surveillance and autonomous-weapons uses
- Supply-chain designation was unprecedented for a US company
New here? Start with this
Anthropic is a US artificial intelligence company best known for Claude, a chatbot and family of AI models that can write, analyse information and assist with software. Its technology is used by businesses and has also attracted interest from government agencies, including the US military.
The Pentagon can use legal powers intended to protect military supply chains from security threats, such as foreign interference or compromised equipment. Being named a supply-chain risk can make it harder for a company to win government contracts and can affect how other customers view it.
The dispute centres on limits Anthropic placed on military use of its models, particularly domestic surveillance and weapons that could operate without direct human control. It also raises wider questions about how far the US government can go in using national-security rules when companies set conditions on the use of their technology.
Both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
Supporters of the ruling argue that government procurement and national-security authorities must not be used to punish a company for setting lawful ethical limits or criticising official policy. They contend that Anthropic had credible concerns about unreliable autonomous weapons and rights-invasive domestic surveillance, and that branding a US firm a supply-chain risk without a genuine security basis could inflict severe commercial and reputational harm while chilling independent oversight.
The case against
Those sympathetic to the Pentagon’s position argue that the military must be able to assess whether essential technology suppliers will reliably support authorised defence missions, particularly where AI capabilities may affect operational readiness. They may believe that restrictions on surveillance or weapons-related uses can create legitimate compatibility and resilience concerns, and that officials need broad discretion to protect military systems—provided such powers are exercised on evidence rather than as retaliation.
Full account
A federal judge in California has ruled that the Trump administration unlawfully retaliated against the artificial intelligence company Anthropic after it refused to remove limits on how the US military could use its models. Judge Rita F. Lin found that the Pentagon’s decision to label the company a supply-chain risk was illegal and without a proper basis. The designation, made in February, was highly unusual: the measure is generally associated with firms regarded as national-security threats, and Anthropic said it was the first US company to be publicly subjected to it.
The dispute arose from negotiations over military contracts for AI services. According to the reports, Defence Secretary Pete Hegseth sought terms allowing the Pentagon to use AI for any lawful purpose. Anthropic said it could support national-security work, but maintained two restrictions: its systems should not be used for mass domestic surveillance or in lethal autonomous weapons capable of selecting and killing targets without human oversight. The company said such uses could threaten rights and democratic values, and argued that its models were not sufficiently reliable for weapons deployment.
After Anthropic declined to change those conditions, the administration escalated pressure and ultimately instructed businesses working with the military to avoid the company. Anthropic sued in March, saying the action damaged its business and violated its free-speech rights. Judge Lin had earlier temporarily blocked the blacklist, observing that the government appeared to be attempting to cripple the company and deter public debate about military AI. In her final ruling, she said national security could not be invoked as a general power to punish critics, while recognising that the Pentagon remained free to choose its own technology suppliers.
Anthropic welcomed the decision and said it wanted to work constructively with government on AI and national security. The reports say the Pentagon and White House had not provided fresh comment to all outlets at the time of publication. The administration had previously argued that private companies should not be able to restrict military action through their terms of service, and had questioned whether Anthropic’s stance gave it undue control over how its technology might be used. The case therefore leaves a broader policy disagreement unresolved: how far AI providers may set binding ethical limits on government customers, and how far the government may respond when it rejects those limits.
Where outlets differ
The BBC report places greater emphasis on the unprecedented nature of applying the supply-chain-risk label to an American company, Anthropic’s claim that its models were not dependable enough for weapons systems, and earlier White House criticism of the firm as 'woke'.
Source 2 gives a fuller account of the contract negotiations, describing the proposed 'any lawful use' terms, Anthropic chief executive Dario Amodei’s public statement, and the Pentagon’s subsequent agreements with other AI companies.
Both reports present Judge Lin’s central finding as unlawful First Amendment retaliation, although Source 2 quotes more extensively from the ruling and describes the decision as unconstitutional.
More coverage
AI Americas Art Business Celebrity Companies Culture Entertainment Technology World