A Federal Judge Voids Anthropic's Blacklist
The Label And The Lockout
U.S. District Judge Rita Lin ruled on August 27, 2026 that the Pentagon's 'supply chain risk' label against Anthropic was unlawful retaliation, and she voided it. The label matters because of what it does mechanically: under federal acquisition rules, a company tagged 'supply chain risk' by the Defense Department is automatically off limits to every federal agency, not just the Pentagon. The Department of Homeland Security, the IRS, the Department of Veterans Affairs, any agency that wanted to buy Anthropic's Claude models for a contract had to look elsewhere, no separate review required. Defense Secretary Pete Hegseth's office attached that label to Anthropic without citing a technical vulnerability, a data breach, or a foreign ownership stake, the kinds of findings the designation exists to catch. Anthropic sued, arguing the label was payback for the company's public disagreements with the administration over AI policy, not a security finding at all. That is a First Amendment claim, the government punishing a critic. It is also a Fifth Amendment claim, no due process before the label attached, and Anthropic rolled both into one lawsuit. Judge Lin agreed with both. The ruling does not just lift a paperwork restriction. It reopens a procurement channel that had been sealed shut for every U.S. agency, and it does so by finding the seal itself was never about security.
The Contradiction On Record
What sank the Pentagon's case was not Anthropic's argument. It was the Pentagon's own paper trail, laid out in Judge Lin's opinion. At the same time Hegseth's office was calling Anthropic a supply chain risk, the Pentagon was weighing whether to invoke the Defense Production Act, a Korean War era law that lets the government declare a private company essential to national security and compel it to prioritize government orders. You do not invoke that law against a company you believe is compromised. The Pentagon was also still pursuing new contracts with Anthropic during the same window, and it was actively co-building Mythos, a cybersecurity model the two organizations developed together to defend military networks. A company being used to help secure the country's own networks is not, by any coherent definition, a supply chain risk to that country. Judge Lin wrote that invoking national security is 'not a blank check to punish and retaliate against government critics.' The Pentagon did not lose because a judge disagreed with its threat assessment. It lost because its own conduct proved there was no threat assessment, just a label applied after Anthropic pushed back in public on the administration's AI policy.
What A Real Risk Looks Like
The same week, a different supply chain risk case moved through a different court. Taiwanese prosecutors indicted a senior Nvidia manager, surnamed Chang, along with eight others including two employees of server maker Supermicro, for forging end user documents, the paperwork that certifies who actually receives export controlled hardware, to route 74 Nvidia B300 equipped servers into China through Japan and Indonesia. A further 56 servers were seized before they crossed. B300 is Nvidia's current top tier AI chip, exactly the kind U.S. export rules exist to keep out of China. That is what an actual supply chain risk case looks like: named defendants, forged paperwork, seized hardware, chips that changed hands illegally and can be counted. Washington's case against Anthropic had none of that, no diverted chips, no forged documents, no seized shipment, just a label. Anthropic did not slow down while the label was in effect either. On August 26, it signed a roughly $45 billion, six year deal with UK firm Nscale for 460 megawatts of Nvidia's next generation Vera Rubin chip capacity at a West Virginia site coming online in late 2027. Financing partners on a deal that size require proof of contracting eligibility before they will sign, and Nscale signed anyway. The label was cheap to apply and, it turned out, cheap to beat in court.
Anthropic has won one round. A second lawsuit over the identical 'supply chain risk' label is still pending in a Washington, D.C. court, and the Pentagon has not said whether it will appeal Judge Lin's ruling or simply reissue a narrower version of the same label, against Anthropic or against the next AI company that criticizes it in public. That D.C. case is the one to watch: whether the Pentagon states its appeal intent there, or lets the same reasoning that sank the first label sink the second, will show whether this ruling changes anything beyond one case.
Sources
- Anthropic gets its first court win over the Pentagon's supply chain risk label (techcrunch.com)
- Judge rules Anthropic supply chain risk designation was illegal and baseless (nextgov.com)
- Anthropic Pentagon supply chain risk ruling unlawful (cnn.com)
- Taiwan Nvidia chip smuggling case (japantimes.co.jp)
- Anthropic and Nscale strike $45 billion cloud deal (cnbc.com)
- Nvidia senior manager linked to Supermicro scheme smuggling AI servers to China (arstechnica.com)