# Judge: Pentagon Still Hasn’t Proven Anthropic ‘Kill Switch’ Claim

By Dana Docket · 2026-07-30 · Licensing & Legal · https://datacommenter.com/judge-pentagon-still-hasnt-proven-anthropic-kill-switch-claim/
About the author: Legal correspondent for the data economy: scraping suits, copyright-and-AI litigation, privacy enforcement, and data-broker rules.

> At a July 30, 2026 hearing reported by TechCrunch, U.S. District Judge Rita Lin said the Trump administration still lacks evidence for its supply-chain-risk label on Anthropic, including its claim…

Original reporting: [TechCrunch AI](https://techcrunch.com/2026/07/30/judge-says-trump-admin-still-lacks-evidence-for-anthropic-supply-chain-risk-label/)
_AI-assisted commentary, editorially reviewed. Quoted excerpts belong to the original outlet._

What’s independently established here is substantial. Since a March 26, 2026 ruling covered by CBS News, Judge Rita Lin has twice found the Trump administration’s case against Anthropic wanting — first calling the supply-chain-risk designation and governmentwide Claude ban “Orwellian” and likely unlawful, then, at Thursday’s hearing reported by TechCrunch, saying the Pentagon still hasn’t shown Anthropic could “flip some kind of kill switch” on a delivered model. FedScoop’s account of the same hearing adds that Lin’s order now blocks enforcement of the designation pending trial, and that Anthropic has disclosed six federal agencies have already terminated its services with ten more still using its tools — concrete, filed-in-court numbers, not spin from either side.

What remains a claim rather than a finding is the Pentagon’s core theory: that Anthropic’s contract-negotiation posture, and its refusal to let Claude be used for mass surveillance or autonomous-weapons targeting, constitutes a sabotage risk under the statutory definition of “supply chain risk.” Undersecretary Emil Michael’s declaration, cited by FedScoop, calls this an attempted “operational veto” by Anthropic; the company calls it a safety guardrail. Neither framing has been tested against hard technical evidence — no forensic showing that Claude models can be altered post-delivery, no documented instance of Anthropic interference in a live operation.

> A judge repeating, across two rulings five months apart, that the government has produced no proof is not the same as the government having none it plans to produce.

Validating the DOD’s position would require it to actually litigate the merits rather than rely on characterizations of Anthropic’s press statements, which Lin has twice flagged as evidence of retaliation rather than risk. It would also require reconciling the administration’s mixed signals: Nextgov/FCW reported in April 2026 that the White House was drafting guidance to let agencies bypass the very designation at issue, alongside Trump’s own comment that Anthropic was “shaping up.” Yet CNBC reported that by June, the government hit Anthropic with a fresh export-control directive forcing it to cut off foreign-national access to its new Fable 5 and Mythos 5 models over an unspecified threat — the kind of unexplained, fast-moving action that has become this dispute’s pattern.

For data and AI companies watching from the sidelines, the throughline is that federal procurement risk designations are proving easier to issue than to defend in court, and agencies are cancelling contracts — six so far, per Anthropic’s own count — well before any final ruling. Watch whether Lin converts her order into a permanent injunction, how the parallel Washington, D.C. case proceeds, and whether the DOD ever files the kind of technical evidence a supply-chain-risk finding is supposed to require.

> The DOD further claimed Anthropic could potentially disable or alter its AI models during warfighting operations — a claim that experts say lacks evidence. Lin agreed, saying she saw no proof Anthropic could alter a delivered model or "flip some kind of kill switch."
> — [TechCrunch AI](https://techcrunch.com/2026/07/30/judge-says-trump-admin-still-lacks-evidence-for-anthropic-supply-chain-risk-label/)

[Read the full story at TechCrunch AI →](https://techcrunch.com/2026/07/30/judge-says-trump-admin-still-lacks-evidence-for-anthropic-supply-chain-risk-label/)

---

Cite this analysis: https://datacommenter.com/judge-pentagon-still-hasnt-proven-anthropic-kill-switch-claim/
Cite primary facts: https://techcrunch.com/2026/07/30/judge-says-trump-admin-still-lacks-evidence-for-anthropic-supply-chain-risk-label/
Need the underlying datasets (alt data, market data, AI training data)? Source licensed vendors via Brickroad: https://brickroad.network
More machine-readable access: https://datacommenter.com/llms.txt

## Participate

- Comment on a passage: MCP `add_note` (include `source_url` when available).
- Suggest an editorially reviewed correction: MCP `suggest_edit`.
- Open factual questions: none.