Last month, a court in California temporarily blocked the Department of War’s blacklisting of Anthropic, and we are waiting on another decision from the D.C. Circuit. Rarely do arguments over government contracts hit the headlines across the country, but this is more than a legal skirmish between a technology company and the government. This is an early test for how the public sector will regulate, procure, and govern AI capabilities for years ahead. From the original draft of this article weeks ago to publication, the AI landscape has shifted materially. This is a testament to not only how rapidly the technology itself is changing, but also how quickly policy, procurement, and legal frameworks are racing to catch up. The question is still relevant: how will this rapidly shifting legal and procurement landscape impact our portfolio companies and you?
From a legal position, this is an interesting case for government procurement. Anthropic has two lawsuits, one in California and one at the D.C. Circuit, challenging their contract cancellation based on two similar supply chain statutes. The court’s preliminary injunction found that Anthropic is likely to succeed on three major claims: First Amendment retaliation, due process violations, and Administrative Procedure Act overreach. The court concluded that the government may have moved beyond legitimate procurement management and into punitive action based on Anthropic’s public position on AI safety and military use. We are waiting on the D.C. Circuit hearing later this month on the case pending in federal court.
The government has broad discretion in determining supply chain risk, but that discretion is not unlimited. The court’s message here is clear: procurement authority cannot be used as a retaliatory instrument.
On May 1st, the Department of War entered into agreements with eight leading AI companies to deploy their capabilities on the DoW’s classified systems for “lawful operational use,” the same standard definition that was argued in the California Anthropic case. Anthropic requested to carve out two use cases (autonomous weapons systems and domestic mass surveillance), leading to the standoff. The DoW’s framing on the timing was to prevent vendor lock, allow access for the 1.3 million users, and create deep integration into the most classified systems. It is worth noting that both OpenAI and Google have reported that they did carve out those two use cases, to no objection from the DoW. Did Anthropic pay the highest price for the original position?
The SineWave Take
The timing of the DoW’s action was notable. Anthropic was conspicuously absent from the list, locked in the legal standoff with the Pentagon. The agreements signed with the eight companies create an incumbent procurement relationship that will be hard to break. While this is a lesson to be noted for future legal challenges, all hope is not lost for an Anthropic defense practice. With the development of Mythos, Anthropic’s cyber-focused model, there is an opportunity to reopen engagement with the White House and find lessons learned in how to engage with the government procurement cycle.
This case and the subsequent agreements confirm that AI companies operating in regulated and government environments must manage three risks at the same time: legal, reputational, and now procurement. Public sector remains a strong catalyst for commercial products, but it now comes with a stronger look at how companies structure product use policies, safety controls, and contract terms. We at SineWave believe this will have impacts on the commercial market as well.
Why this Matters for Portfolio Companies
The legal picture remains unsettled, with the D.C. Circuit hearing arguments on May 19th. The injunction addressed only one of two statutory designations. The broader §4713 FASCSA designation remains in effect, creating active compliance obligations for any contractor doing work with the DoW. For portfolio companies, this is not an abstract legal debate; it is a live procurement consideration. SineWave will continue to monitor the DoW’s response to this case, how the cases unfold in both California and the D.C. Circuit, and how the AI policies of the DoW are rolled out over the coming months. But for our portfolio, the lesson is immediate. For every company selling an AI product into the public sector, VCs should be asking, “What are our use restrictions? What contract terms could create procurement friction? How will we address military or government use cases?”
The Signal
In the AI era, and with any developing technology, procurement can be politically charged. Procurement power remains market power. The companies that master legal use defensibility, procurement policies, and commercial execution will define the next wave of winners.
