A federal judge in Virginia has signaled serious concern about the Department of Defense’s decision to effectively shut Anthropic out of a massive cloud computing contract — a move that could reshape how the U.S. government procures artificial intelligence and which companies get to build the infrastructure undergirding national security.
The case centers on the Pentagon’s Joint Warfighting Cloud Capability contract, known as JWCC, a multi-billion-dollar program that provides cloud services across all classification levels to the Department of Defense. Amazon Web Services and its competitors have long jockeyed for position on the contract. But Anthropic, the San Francisco–based AI company behind the Claude family of models, found itself caught in a contractual dispute that has now escalated into a legal battle with implications far beyond a single procurement.
Here’s the short version: Anthropic had a deal with AWS to offer its AI models through the JWCC contract. The Pentagon intervened, issuing guidance that effectively barred Anthropic’s models from being accessed through the contract. Anthropic sued. And now a judge at the U.S. Court of Federal Claims has used the word “troublesome” to describe the government’s conduct.
That word matters.
As WIRED reported, Judge David Tapp didn’t mince his assessment during a hearing on Anthropic’s request for a temporary restraining order. Tapp said the government’s actions were “troublesome” and questioned whether the Pentagon had improperly targeted a specific company. The judge stopped short of immediately granting the restraining order but made clear that Anthropic’s arguments had merit and that the case would proceed with urgency.
The dispute has its roots in how the JWCC contract is structured. Unlike the Pentagon’s earlier, ill-fated JEDI contract — a winner-take-all approach that collapsed under legal challenges from Microsoft and Amazon — JWCC was designed as a multi-vendor vehicle. AWS, Microsoft Azure, Google Cloud, and Oracle all hold positions on it. The idea was to avoid the kind of monopolistic arrangement that doomed JEDI and instead let military agencies pick from multiple providers.
Anthropic doesn’t hold a JWCC contract directly. It operates as a subcontractor, offering its AI models through AWS’s cloud infrastructure. This is a common arrangement in government IT: prime contractors win the big vehicle contracts, then bring in specialized technology partners to deliver specific capabilities. For months, defense and intelligence agencies had been accessing Claude models this way.
Then the Pentagon stepped in.
According to court filings reviewed by multiple outlets, the Defense Department issued contracting guidance that restricted how AI models from companies like Anthropic could be offered under JWCC. The practical effect was to block Anthropic’s models from being purchased through the contract. Anthropic argues this was done without adequate justification and that it amounts to an arbitrary restraint of trade that harms both the company and the government agencies that want to use its technology.
The timing is conspicuous. The Trump administration has been reshaping federal technology procurement at a rapid pace, with close ties to certain Silicon Valley figures influencing which companies gain or lose favor. Elon Musk’s Department of Government Efficiency — DOGE — has been involved in reviewing and canceling government contracts across agencies. And the broader political dynamics around AI procurement have become increasingly fraught, with companies like Palantir, Anduril, and others competing aggressively for defense AI dollars.
Anthropic occupies an unusual position in this contest. The company was founded by former OpenAI executives Dario and Daniela Amodei, and it has positioned itself as the safety-focused alternative in the AI race. It has also been building significant government business, particularly through its partnership with AWS, which invested billions in the company. Anthropic’s Claude models have been gaining traction with intelligence and defense users who value both the models’ capabilities and the company’s approach to safety testing and red-teaming.
But safety-mindedness doesn’t necessarily translate into political favor. And Anthropic lacks the deep defense establishment relationships that companies like Palantir have cultivated over decades.
Judge Tapp’s comments carry weight because the Court of Federal Claims is the primary venue for government contract disputes. When a judge there calls a procurement action “troublesome,” it’s a signal to both the contracting community and the broader defense establishment that something may have gone wrong in the process. It doesn’t guarantee Anthropic will win. But it dramatically increases the pressure on the Pentagon to justify its decision.
The government’s defense, as outlined in court filings, appears to rest on the argument that it has broad discretion in how it manages the JWCC contract and that the restrictions on subcontractor AI models were a reasonable exercise of that authority. Pentagon officials have also suggested that security and interoperability concerns justified the decision.
Anthropic’s lawyers pushed back hard. They argued that the company’s models had already been approved for use on classified networks through AWS’s infrastructure, that multiple defense agencies were actively using Claude, and that the Pentagon’s sudden reversal caused immediate and irreparable harm. The company also pointed to what it characterized as procedural irregularities — suggesting the decision wasn’t made through normal acquisition channels.
The financial stakes are enormous. Government AI spending is projected to surge in the coming years, with the defense and intelligence communities representing the single largest pool of potential revenue for AI companies. Being locked out of JWCC doesn’t just mean losing current business. It means losing the foothold that leads to future contracts, pilot programs, and the kind of institutional adoption that compounds over time.
For AWS, the dispute is also consequential. Amazon has invested roughly $4 billion in Anthropic and has been marketing Claude models as a key differentiator for its government cloud offerings. If the Pentagon can unilaterally decide which subcontractor AI models are available through JWCC, it undermines AWS’s ability to offer its full range of capabilities — and it raises questions about the value of the JWCC contract itself.
There’s a broader principle at play too. The multi-vendor structure of JWCC was supposed to promote competition and give agencies flexibility. If the Defense Department can selectively restrict which technologies are available through the contract without transparent justification, it defeats the purpose of the multi-vendor approach. Other AI companies watching this case — and there are many — are acutely aware that what happens to Anthropic today could happen to them tomorrow.
So where does this go from here? Judge Tapp indicated he would rule quickly on the temporary restraining order, and the case is likely to move through expedited proceedings given the national security dimensions. If the judge grants injunctive relief, the Pentagon would be forced to restore Anthropic’s access through JWCC while the case is litigated. If not, Anthropic faces the prospect of watching its government business erode in real time as agencies are forced to find alternatives.
The case also feeds into a growing tension between the current administration’s stated goal of American AI dominance and its actual procurement practices. President Trump has repeatedly said the United States must lead the world in AI. Executive orders have called for accelerating AI adoption across the federal government. Yet here is the Pentagon apparently restricting access to one of America’s most capable AI systems — made by an American company, running on American cloud infrastructure.
That contradiction hasn’t gone unnoticed on Capitol Hill. Several members of Congress with oversight of defense technology programs have been quietly monitoring the dispute, according to people familiar with the matter. If the case reveals that political considerations — rather than legitimate security or technical concerns — drove the Pentagon’s decision, it could trigger congressional scrutiny of AI procurement practices more broadly.
Anthropic, for its part, has tried to keep the dispute focused on legal and procedural grounds rather than turning it into a political fight. The company’s public statements have emphasized its commitment to serving government customers and its belief that the procurement process should be fair and transparent. Behind the scenes, though, the company has been marshaling significant legal resources and reaching out to allies in Congress and the national security community.
The outcome of this case will send a signal well beyond the courtroom. It will tell AI companies whether the federal procurement system can be trusted to operate on merit, or whether access to the government’s most lucrative contracts depends on factors that have nothing to do with technology. It will tell defense agencies whether they have genuine freedom to choose the best tools for their missions, or whether those choices will be made for them by political appointees with their own agendas.
And it will tell the world whether the United States is serious about maintaining its lead in artificial intelligence — or whether it’s willing to hobble its own companies for reasons that, as a federal judge has now suggested, are troublesome indeed.
WebProNews is an iEntry Publication