US Court Temporarily Backed Anthropic in Dispute with Pentagon Over Claude Restrictions

Anthropic has won a preliminary victory in its dispute with the Pentagon: a federal court in California has temporarily halted measures by which U.S. authorities attempted to restrict the company's work with the government. This is not a final ruling in the case, but a legal pause that allows Anthropic to contest the administration's actions without immediate damage to its business.

Why the court intervened

Judge Rita Lin issued a temporary injunction while the U.S. District Court for the Northern District of California considers the lawsuit on its merits.

At the same time, she delayed the effectiveness of the ruling for one week to give the government time for appeals. For Anthropic, this is the first clear success in a public conflict that, over the course of several weeks, has grown from a contractual dispute into a political and legal case at the federal level. According to the company, the Pentagon and the Trump administration went far beyond the usual right of a customer to refuse a vendor.

Anthropic argues that authorities did not simply terminate cooperation but attempted to punish it for its public position on AI safety. In the lawsuit, the company invokes violations of the First Amendment, since the dispute concerns statements on a matter of public significance: whether it is permissible to use generative models for autonomous weapons and mass surveillance.

The root of the conflict

The root of the conflict lies in Anthropic's red lines for Claude. The company refused to allow the use of the model in fully autonomous lethal systems where the final decision to strike a target is made by a machine without human involvement. The second restriction concerns mass surveillance of citizens within the country.

The Pentagon, by contrast, insisted on language that would allow the model to be used for all lawful purposes, regardless of additional corporate restrictions. After negotiations failed, the dispute escalated rapidly. Authorities designated Anthropic as a supply chain risk and then demanded it cease using its technologies in government agencies.

According to case materials, the company sees such steps not as a security measure but as a pressure tactic designed to force it to soften its own rules.

• Anthropic refused to remove restrictions on autonomous weapons
• The company refused to open Claude for mass domestic surveillance
• The administration demanded an end to the use of Anthropic's technologies in government agencies
• The Pentagon designated the company as a supply chain risk
• Anthropic stated this threatens it with multimillion-dollar damages and reputational harm

What the judge found

In her preliminary decision, Judge Rita Lin essentially agreed with Anthropic's main thesis: the government probably went further than the law allows. Court documents state that designating the company as a supply chain risk may have been arbitrary and poorly justified. Lin separately noted that authorities had a simpler path: if the Defense Department was not satisfied with the conditions for using Claude, it could simply stop buying the product and switch to another vendor.

"This looks like an attempt to paralyze Anthropic."

This moment became one of the key points for the court. If the government's goal was to protect its operational interests, it did not have to impose such broad restrictions. The court also noted that public criticism of the government's position and discussion of AI risks fall under freedom of speech protections.

In other words, the case now looks not like an ordinary procurement dispute but like a possible retaliatory measure against the company for its position. The ruling does not require the Pentagon to use Claude and does not prohibit the department from seeking another contractor. The ban concerns broader punitive measures: blacklisting, supply chain threat status, and directives that effectively cut the company off from working with the government before the final review of the lawsuit.

What this means

This dispute sets an important precedent for the entire AI model market. The court has shown that even in the sphere of national security, it will be difficult for the state to force developers to drop basic restrictions through pressure on contracts and reputation. For Anthropic, this is not yet a victory in the case, but a strong signal: the court is ready to take seriously the right of companies to draw their own red lines in AI use.