The U.S. government cannot blacklist Anthropic. At least not yet.
A federal judge in the Northern District of California handed Anthropic a significant legal victory on Thursday, granting a preliminary injunction that blocks the Pentagon from labeling the AI company a “supply chain risk” — a designation that would have barred federal contractors from using Claude and other Anthropic technology. Judge Rita Lin ruled that the Department of War lacked the legal authority to issue such a designation after Anthropic sued the government over what it called a retaliatory and politically motivated move.
The ruling is more than a legal footnote. It signals that courts are willing to examine the government’s use of national security designations as tools in what has become an increasingly bitter dispute between the Pentagon and one of the most prominent AI safety-first companies in the world.
What the Pentagon Tried to Do
The backstory is worth unpacking. Earlier this year, the Pentagon moved to place Anthropic on a list that would effectively cut the company off from federal contracts and discourage contractors from working with it. The administration pointed to concerns about Anthropic’s refusal to sign certain defense contracts and what it described as the company’s unwillingness to align its AI development priorities with national security needs.
Anthropic pushed back hard. The company, which has built its brand around constitutional AI and safety-focused development, filed suit arguing that the designation was retaliatory — punishment for the company’s public stance on AI safety and its reluctance to pursue aggressive defense partnerships. The injunction filing detailed what Anthropic characterized as a coordinated campaign to sideline a competitor that the administration viewed as insufficiently cooperative.
The Seven-Day Clock
The injunction itself comes with a significant caveat. The ruling is stayed for seven days from March 26, giving the government time to file an emergency appeal. That means the practical effect of the block does not kick in until approximately April 2, 2026. If the government pursues an appeal, the legal battle could stretch well into the spring, creating a period of continued uncertainty for federal agencies and contractors that had begun restructuring their AI procurement in anticipation of the ban.
This timeline matters. Federal agencies and contractors have been watching this story closely, knowing that the outcome could reshape how they source AI technology. Companies that were already diversifying away from Anthropic are now in a holding pattern. Those that had quietly increased their reliance on Claude are watching the calendar.
A Wider Pattern in the AI Policy Landscape
This case sits at the intersection of several overlapping tensions in the AI world right now. The administration has been expanding its use of executive power to shape the AI industry, from export controls on chips to pressure on model developers around safety standards and defense applications. The Pentagon has pursued aggressive contracting strategies that favor companies willing to integrate deeply with military systems.
Anthropic has occupied a distinct position. It has worked with the defense sector — notably through its partnership with Palantir on the Maven smart program — but has been notably more conservative than OpenAI about the scope of its military commitments. The company has also been more publicly vocal about AI safety risks, positioning itself as the thoughtful counterweight to what it views as reckless speed-first competition.
That positioning has won Anthropic a devoted following among AI safety researchers and enterprise customers who value its principles. But it has also made the company a target in an administration that views hesitation around defense work as tantamount to disloyalty. Thursday’s ruling is a check on that impulse, though not a final verdict.
What Comes Next
The legal process will now move into its next phase. The government can appeal to the Ninth Circuit, a venue known for scrutinizing executive action carefully. Anthropic will need to continue building its case for why the supply chain designation was improper. Meanwhile, the broader question of how the U.S. government should relate to AI companies that set conditions on how their technology gets used is only going to become more pressing as AI systems grow more capable and more embedded in critical infrastructure.
The judge used the word “preliminary” deliberately. This is not the end of the story. It is a pause — a legal intervention that gives Anthropic room to continue operating in the federal market while the fight plays out in full. But it is also a meaningful signal that the courts are not simply deferring to the executive branch when AI companies allege that national security labels are being weaponized for commercial or political purposes.
For the AI industry broadly, the case is becoming a reference point for how far government power can stretch in directing the development and deployment of advanced AI systems. Whatever happens next in the courts, the precedent being set here will outlive this particular dispute.