# A judge struck down the Anthropic blacklist. The Pentagon can still walk away.

URL: https://www.thedeepfeed.ai/posts/2026-08-28-court-strikes-down-anthropic-blacklisting/
Category: Policy
Published: 2026-08-28
Author: the-deep-feed
Tags: policy, anthropic, pentagon, courts, government-contracting, first-amendment
Kind: news

> Judge Rita Lin ruled on August 27 that the administration's February blacklisting of Anthropic was illegal retaliation, vacating the supply-chain-risk label and the contractor boycott while leaving the Pentagon free to stop buying Claude.

## TL;DR

- On **August 27**, Judge **Rita F. Lin** of the Northern District of California ruled in *Anthropic PBC v. U.S. Department of War* that the February blacklisting was First Amendment retaliation and a due-process violation, and entered final judgment.
- The order vacates the **10 U.S.C. § 3252** supply-chain-risk designation and Secretary **Pete Hegseth**'s ban on any military contractor doing business with Anthropic, and permanently enjoins the agencies that carried out the directive.
- It also says the Pentagon may still stop buying Claude. As of **August 28** the government had not commented; the AP reported it is expected to fight, and a separate designation case is still pending at the **D.C. Circuit**.

On August 27, Judge Rita F. Lin of the Northern District of California ruled that the Trump administration's February blacklisting of **Anthropic** was illegal. In a [59-page opinion](https://storage.courtlistener.com/recap/gov.uscourts.cand.465515/gov.uscourts.cand.465515.250.0.pdf) in *Anthropic PBC v. U.S. Department of War* (No. 3:26-cv-01996-RFL), she found the government's actions were unconstitutional retaliation for Anthropic's public criticism, denied the company due process, and misused a procurement statute written for saboteurs. A separate [order of final relief](https://storage.courtlistener.com/recap/gov.uscourts.cand.465515/gov.uscourts.cand.465515.251.0.pdf) vacated the designation and the contractor boycott, and the court entered final judgment the same day. The ruling also makes a point the coverage has mostly skipped. It separates two powers the administration had used as one: the right to stop buying from a vendor, which the court left intact, and the power to label a domestic vendor a national-security threat over its contract terms and public statements, which it struck down.

# What the government did, and why

The opinion's undisputed facts start with a contract clause. In fall 2025 the Department of War told Anthropic it had to drop its usage restrictions and accept use of Claude "for all lawful uses." Anthropic agreed to remove most of them but refused to drop two: lethal autonomous warfare and mass surveillance of Americans. At a February 24 meeting, Secretary Pete Hegseth called Claude's capabilities "exquisite," said the department "had many other vendors to choose from," and set a 5 p.m. deadline on February 27.

At 3:47 p.m. that day, before the deadline, President Trump posted on Truth Social that he was "directing EVERY Federal Agency" to "IMMEDIATELY CEASE all use of Anthropic's technology," with a six-month phase-out for agencies such as the Department of War. A little over an hour later Hegseth posted that he was directing the department to designate Anthropic a supply-chain risk and that "no contractor, supplier, or partner that does business with the United States military may conduct any commercial activity with Anthropic." A formal designation under 10 U.S.C. § 3252 followed on March 3. Agencies including GSA, OPM, the State and Treasury departments and the Nuclear Regulatory Commission cut off access within days. [Reuters](https://www.thestar.com.my/tech/tech-news/2026/08/28/us-judge-blocks-pentagon039s-anthropic-blacklisting) reported it was the first time a U.S. company had been publicly designated under that statute, which was written to protect military systems from foreign sabotage.

The written justification was a four-page memo by Under Secretary Emil Michael, dated March 2. It said Anthropic's risk level rose when it "insisted on terms of service that would constrain the DoW beyond what is in the law" and began "engaging in an increasingly hostile manner through the press." Anthropic sued on March 9. Lin issued a preliminary injunction on March 26, and the government appealed it to the Ninth Circuit without seeking an emergency stay.

# What the court held and ordered

The opinion is blunt about the record. "The record is slim," it begins. The government had "backed away from" its claim that Anthropic could reach into deployed models; Anthropic "undisputedly lacks any such access," and the government conceded Claude is "no riskier to the national security than any other 'black box' artificial intelligence model." What remained was trust, and Lin held that the government could not ground "sweeping penalties based principally on Anthropic's critique of the Administration's views."

On the statute, she read § 3252's definition of supply-chain risk, "the risk that an adversary may sabotage, maliciously introduce unwanted function, or otherwise subvert" a national security system, as "directed at covert acts or hacks, not overt, public positions regarding contract terms." The government conceded there was no statutory basis for Hegseth's ban on contractors doing business with Anthropic. The process also failed on its own terms. Nothing in the record discussed the "less intrusive measures" the statute requires the Secretary to consider. The risk assessment came from Michael, the official negotiating the contract, rather than the intelligence under secretary the department's own regulations name. The factual basis was prepared one business day after Hegseth's post. And on March 4, the day after the designation, Michael emailed Anthropic: "I think we are very close here."

Anthropic did not win everything. Lin rejected its separation-of-powers claim against the presidential directive and ruled for HHS, Commerce, Veterans Affairs, the SEC and NASA on one statutory claim, finding they had taken no relevant action or only interim steps. The relief is still broad:

| What the order does | Scope |
|---|---|
| Declares the actions violate the First Amendment and Fifth Amendment due process | All participating defendants |
| Permanently enjoins implementing or enforcing the directives, and directs agencies to rescind guidance issued under them | All defendants except the non-participating ones (NEA, SSA, the Fed board, the Executive Office of the President) |
| Vacates the § 3252 designation and the contractor-boycott order | Department of War and Hegseth |
| Vacates agency actions taken to carry out the presidential directive | DoW, State, OPM, NRC, Treasury, DHS, Energy, FHFA, GSA |
| Leaves the department free to switch AI vendors | Paragraph 14 |

Lin also denied the government's request to put the injunction on hold for seven days, noting it had complied with the preliminary injunction "for more than five months" without identifying any harm.

# The government's response so far

As of Friday, the administration had said nothing on the record. Reuters and Al Jazeera reported that the Pentagon did not immediately respond to requests for comment, the [AP](https://federalnewsnetwork.com/artificial-intelligence/2026/08/judge-says-pentagons-measures-against-anthropic-were-illegal-and-baseless/) said the White House had not responded, and it reported that "the government is expected to fight the ruling." In court the Justice Department had argued that the designation stemmed from Anthropic's refusal of contract terms rather than its views, and [CBS News](https://www.cbsnews.com/news/judge-rules-trump-administration-illegally-punished-ai-firm-anthropic/) reported it described agencies' decisions to drop Claude as "internal IT housekeeping." Anthropic said it welcomed the ruling and remained "focused on working productively with the government to harness AI for our national security."

The case is not over. The Ninth Circuit appeal of the March injunction (No. 26-2011) has been stayed at both parties' request pending a ruling from the D.C. Circuit in a related case (No. 26-1049). That case, as Reuters describes it, concerns a separate Pentagon supply-chain designation that could lead to Anthropic's exclusion from civilian government contracts, and it remains undecided.

# What it means for labs selling to Washington

For other AI companies, the ruling draws a line rather than settling the commercial question. Hegseth's January 9 AI strategy memo, cited in the opinion, requires an "any lawful use" provision in "any DoW contract through which AI services are procured," so every lab bidding for defense work faces the term Anthropic refused. Lin's order does not make that term negotiable. It says a lab can refuse it, argue about it in public, and lose the contract, without being branded an adversary. In her words, "An IT vendor does not become a potential adversary of the United States whenever it asks probing questions or stubbornly insists on particular contracting terms."

The stakes the opinion records are not abstract. Anthropic told the court that if the designation stood, its defense-related revenue would fall by "50–100 percent" and its 2026 revenue by "multiple billions of dollars." More than 100 enterprise customers had asked about the fallout, and one partner moved an FDA contract to a rival model. Trade associations for government contractors told the court their members faced "a compliance crisis." For those contractors the practical result is narrower: the instruction not to do "any commercial activity" with Anthropic no longer has legal force, though the government may yet appeal the order that vacated it.

On [Hacker News](https://news.ycombinator.com/item?id=49473522), where the ruling drew a 437-comment thread, readers split over whether it would stick. One commenter, andrewla, read its logic this way: ["The evidence being weak does not invalidate the decision; as the court notes, the government gets substantial deference on matters of national security. The invalidation comes from the fact that this is clearly retaliatory behavior for speech"](https://news.ycombinator.com/item?id=49478376).

That reading matches the opinion, and it defines what the ruling protects. The same government restricted Anthropic's models through [a Commerce export letter in June](/posts/2026-06-27-government-joined-the-model-release/), and the opinion cites April news reports that the NSA and NASA had begun deploying Anthropic's Mythos model. A court can stop the state from punishing a vendor for what it says. It cannot make the state a customer. As Lin put it, "The empty invocation of national security is not a blank check to punish and retaliate against government critics."

## Sources

- [N.D. Cal. — Anthropic PBC v. U.S. Department of War, No. 26-cv-01996-RFL, Order on cross motions for summary judgment, Dkt. 250 (Aug 27, 2026)](https://storage.courtlistener.com/recap/gov.uscourts.cand.465515/gov.uscourts.cand.465515.250.0.pdf)
- [N.D. Cal. — Anthropic PBC v. U.S. Department of War, Order of Final Relief, Dkt. 251 (Aug 27, 2026)](https://storage.courtlistener.com/recap/gov.uscourts.cand.465515/gov.uscourts.cand.465515.251.0.pdf)
- [Reuters via The Star — US judge blocks Pentagon's Anthropic blacklisting (Aug 27, 2026)](https://www.thestar.com.my/tech/tech-news/2026/08/28/us-judge-blocks-pentagon039s-anthropic-blacklisting)
- [AP via Federal News Network — Judge says Pentagon's measures against Anthropic were 'illegal and baseless' (Aug 28, 2026)](https://federalnewsnetwork.com/artificial-intelligence/2026/08/judge-says-pentagons-measures-against-anthropic-were-illegal-and-baseless/)
- [CBS News — Judge rules Trump administration illegally punished AI firm Anthropic (Aug 28, 2026)](https://www.cbsnews.com/news/judge-rules-trump-administration-illegally-punished-ai-firm-anthropic/)
- [Hacker News — Judge rules Trump administration's blacklisting of Anthropic was illegal (650 points, 437 comments)](https://news.ycombinator.com/item?id=49473522)

---

Canonical: https://www.thedeepfeed.ai/posts/2026-08-28-court-strikes-down-anthropic-blacklisting/
Site: https://www.thedeepfeed.ai
Full corpus: https://www.thedeepfeed.ai/llms-full.txt