For most of 2026 the most consequential AI policy fight in the US wasn’t in Congress. It was between one AI lab and two federal departments, and it is still in the courts.

The dispute

Anthropic had held a $200 million Pentagon contract since July 2025. According to CNBC, the Defense Department wanted to be able to use Claude for all lawful purposes. Anthropic wanted assurances that its models would not be used in fully autonomous weapons or for domestic mass surveillance.

In March 2026 the department designated Anthropic a “supply chain risk”, which bars military use of its models. Defense Secretary Pete Hegseth accused the company of trying to seize veto power over military operational decisions, CNBC reported.

The model freeze

In June the conflict reached Anthropic’s commercial customers:

  • June 12: A Commerce Department export directive, citing national security authorities, required Anthropic to stop foreign nationals, including its own employees, from accessing Fable 5 and Mythos 5. Anthropic said it couldn’t verify nationality in real time, so it disabled the models for all customers.
  • June 26: Commerce Secretary Howard Lutnick approved Mythos 5 for selected companies and agencies. OpenAI, under a similar government request, limited its own launch to trusted partners the same day, according to CNBC.
  • June 30: The controls were lifted. Fable 5 returned worldwide on July 1.

Why Commerce acted is disputed; some coverage has linked it to a possible jailbreak of Fable 5, which we have not been able to confirm.

The court rulings

  • August 27: US District Judge Rita Lin in San Francisco ruled one designation unlawful, finding it was First Amendment retaliation. She wrote that it appeared based on a desire to make a public example of the company.
  • September 25: A divided D.C. Circuit panel upheld a second designation, made under a different statute. Judge Gregory Katsas wrote the opinion, Judge Neomi Rao joined and Judge Karen Henderson dissented. The ruling is on hold while Anthropic seeks further review.

Why it matters

  • A model you build on can disappear by government order. For about two and a half weeks in June, every developer using Anthropic’s top models lost access with no notice. Multi-provider fallbacks stopped being a nice-to-have.
  • New tools, first uses. These were the first known uses of export controls and supply-chain designations against a US lab’s deployed model.
  • The core question is unresolved. Whether an AI company can restrict how the military uses its products is now a live legal question, and the split between the two courts makes further appeals likely.

Anthropic’s IPO filing in October lists the dispute among its risk factors.

Sources

  1. CNBC: Judge blocks Pentagon blacklist of Anthropic (Aug 27, 2026)
  2. CNBC: Appeals court rules on Pentagon’s Anthropic AI risk designation (Sep 25, 2026)
  3. CNBC: Anthropic says Trump administration has lifted export controls on Claude Fable 5 and Mythos 5 (Jun 30, 2026)
  4. Anthropic: Claude Mythos

We link every claim to the reporting or official statement it comes from. Where a detail rests on one outlet’s reporting, we say so. Spotted an error? Reply to any newsletter and we’ll correct it.