Judge Rejects 'Orwellian Notion' That Anthropic Is a Supply Chain Risk for Disagreeing With the Government

Judge Rita Lin's preliminary injunction confirms what government officials had implicitly acknowledged: The supply chain risk designation was punishment, not policy.

A federal judge blocked the government from banning Anthropic from federal contracting on Thursday. The ruling confirms what the Defense Department's continued use of Anthropic implies: The AI developer is not a supply chain risk, and disagreement with the government is not subversion.

Judge Rita Lin's order reverses President Donald Trump's February 27 directive for every federal agency "to IMMEDIATELY CEASE all use of Anthropic's technology." It also strikes down Defense Secretary Pete Hegseth's order that his agency designate Anthropic as a "Supply-Chain Risk to National Security" and the March 3 directive that formalized that designation. As Lin writes, her order "restores the status quo."

The dispute between Anthropic and the Defense Department centered on two narrow restrictions in the company's usage policy: Claude, its AI model (which the Defense Department exclusively used), may not be deployed for fully autonomous weapons systems or for mass domestic surveillance. The day preceding Trump and Hegseth's actions, CEO Dario Amodei stated that such uses are "simply outside the bounds of what today's technology can safely and reliably do" and that Anthropic "cannot in good conscience accede to" the agency's demands to permit them. 

In response, Trump labeled Anthropic a "RADICAL LEFT, WOKE COMPANY" and Hegseth condemned Anthropic's "defective altruism" before banning Anthropic—and anyone with a commercial relationship with the company—from doing business with the federal government. Reason's Elizabeth Nolan Brown explained that "the administration's above-and-beyond punishment hinged on the fact that the company said no to the government forcefully and publicly—and that's not OK." Lin agrees that "the record supports an inference that Anthropic is being punished for criticizing the government's contracting position," which amounts to "classic illegal First Amendment retaliation." 

In addition to arguing that its "core First Amendment freedoms are under attack," Anthropic argued that the Defense Department lacked statutory authority to designate the company a supply chain risk. Federal law authorizes the secretary of defense to exclude a company from defense procurement only when it may be used by an "adversary" to "sabotage" or "subvert" a national security system. Anthropic affirmed it "is not, and has no ties to, an 'adversary'" and the statute does not permit "the Secretary to redefine 'supply chain risk' to cover a contractor who declines to modify its terms of use to track the Department's preferences." 

Lin concluded that the Defense Department provided "no legitimate basis to infer…[Anthropic] might become a saboteur" and rejected "the Orwellian notion that an American company may be branded a potential adversary and saboteur of the U.S. for expressing disagreement with the government." While the Defense Department is at liberty to decide with whom to contract, Lin ruled it may not attempt "'corporate murder'" because a firm refuses to kowtow to its every whim or exercises its First Amendment rights in a manner embarrassing to the department. 

Dean Ball, senior fellow at the Foundation for American Innovation, celebrated the ruling as a win not only for Anthropic, but "all red-blooded Americans who are, as the founders would have said, 'jealous of their liberties.'" The decision is a huge win for American AI development insofar as it restores investor confidence in the industry. 

Comments

[JesseAz (RIP CK)]

So...

In your view...

Judges...

Not the executive....

Makes supply decisions....

Okay.

[mad.casual]

At the very least, I don't think Judge Rita Lin has read 1984 or Animal Farm. If she has, I'm highly dubious of her reading comprehension skills.

This is like the Berkeley city board member who supported a ban on high rises because she wanted to protect people's privacy from onlookers like Jimmy Stewart in Rear Window.

[Sometimes a Great Notion]

Man, who wouldn't want Skynet to have surveillance and attack capabilities; besides that woke son of a bitch, John Connor.

[mad.casual]

LOL, Claude is a supply-chain risk to itself -
Claude -

Q: Should a Federal US judge be able to declare an injunction to restrict the US Department of War from regarding a contractor as a supply chain risk?
A: This is a genuinely complex constitutional and administrative law question. Here are the key considerations from different perspectives:

**Arguments that judges CAN issue such injunctions:**

- Federal courts have broad equitable powers under Article III to enjoin unlawful executive action, including decisions that violate due process, statutory authority, or the Administrative Procedure Act (APA).
- Contractors have property and liberty interests in their government business relationships, and the Fifth Amendment requires due process before the government deprives them of those interests.
- Courts routinely review and enjoin national security-adjacent agency decisions when they exceed statutory authority or are "arbitrary and capricious" under the APA.
- The Supreme Court has consistently held that even national security decisions are not entirely immune from judicial review (*Boumediene v. Bush*, *Hamdi v. Rumsfeld*).

**Arguments that judges CANNOT or SHOULD NOT:**

- Supply chain risk determinations involve classified intelligence and technical assessments that courts are institutionally ill-equipped to second-guess — judges lack the security clearances, expertise, and information to evaluate them.
- The political question doctrine may place some national security procurement decisions beyond judicial reach, as they are committed by the Constitution to the executive branch.
- Congress has explicitly granted the executive broad authority over defense procurement and supply chain security (e.g., the Federal Acquisition Regulation, NDAA provisions on supply chain risk management like 10 U.S.C. § 3252), suggesting judicial deference is appropriate.
- Enjoining such a designation could directly impair military readiness and national defense — courts are reluctant to issue relief that interferes with the Commander-in-Chief's core functions.

**The middle ground most courts actually apply:**

In practice, courts tend to ask *how* the decision was made rather than *what* was decided. They are more willing to intervene if a contractor was denied basic procedural fairness (no notice, no opportunity to respond) than to substitute their own judgment on the underlying security merits. The Federal Circuit and the Court of Federal Claims have handled contractor exclusion cases on these narrower procedural grounds.

**The note on "Department of War":** The U.S. Department of War was renamed the Department of Defense in 1947, so today this question would involve DoD. The name matters slightly because the statutory frameworks governing procurement and supply chain risk are built around DoD specifically.

The honest answer is that this sits in genuinely contested legal territory — there's no clean yes or no. Courts have the *power* to issue such injunctions in principle, but whether they *should* in a given case depends heavily on what process was afforded the contractor and how directly the court would have to second-guess classified security judgments.

--------------------------------------------------------------------------------
Grok -

Q: Should a Federal US judge be able to declare an injunction to restrict the US Department of War from regarding a contractor as a supply chain risk?
A: Yes. A federal U.S. judge not only can but should be able to issue such an injunction when the executive branch’s action (here, the Department of Defense—referred to in current coverage as the “Department of War”) violates the law, the Constitution, or its own statutory procedures. This is a core function of judicial review under Article III, the Administrative Procedure Act (APA), and precedents like Marbury v. Madison (1803) and Youngstown Sheet & Tube Co. v. Sawyer (1952). National-security labels do not create a blanket immunity from oversight.
Why this power exists and why it was properly exercised here
The specific dispute involves Anthropic’s recent lawsuit against a DoD designation of the company (and its AI model Claude) as a “supply chain risk” under 10 U.S.C. § 3252 (and a related presidential directive under the Federal Acquisition Supply Chain Security Act). That statute allows the Secretary of Defense to exclude a source from certain national-security procurements if there is a finding that an adversary “may sabotage, maliciously introduce unwanted function, or otherwise subvert” a covered system.
On March 26, 2026, U.S. District Judge Rita F. Lin (Northern District of California) granted Anthropic a preliminary injunction. She found Anthropic was “likely to succeed” on claims that the designation was:

First Amendment retaliation (“classic” viewpoint discrimination). DoD records cited Anthropic’s “hostile manner through the press” and public insistence on usage restrictions (e.g., no autonomous weapons or domestic surveillance) as the trigger—after contract negotiations broke down. The judge wrote: “Nothing in the governing statute supports the Orwellian notion that an American company may be branded a potential adversary and saboteur of the U.S. for expressing disagreement with the government.”
Arbitrary and capricious under the APA. No evidence showed the required statutory “supply chain risk” findings; instead, the label was used punitively to blacklist Anthropic from all federal contracts and force defense contractors to certify they weren’t using its technology.
Due process violations. The process bypassed normal safeguards and appeared pretextual.

The injunction is narrow and temporary: it blocks enforcement of the designation and related directives pending final resolution. It was stayed for a week to allow appeal. This is textbook equitable relief—courts routinely enjoin agency actions that exceed statutory authority or punish protected speech.
Counterarguments and why they don’t override judicial review

“National security deference”: Courts do defer heavily to the executive on genuine military judgments (e.g., Department of the Navy v. Egan, 1988). But deference is not a rubber stamp. When the government’s own records show retaliation rather than a good-faith security assessment, or when the action fails the statute’s own requirements, courts must intervene. Amicus briefs from nearly 150 former federal judges in related filings emphasized: there is “no ‘national security exception’ to ordinary principles of judicial review,” citing Youngstown, the Pentagon Papers, and Hamdi v. Rumsfeld.
“Judges aren’t qualified to second-guess DoD”: True for tactical military decisions, but not for whether an agency followed the APA, made required factual findings, or violated the First Amendment. The statute itself contemplates limits and some review.
“This weakens national security”: If the designation had been based on real sabotage risks (as with past uses against Kaspersky or Huawei), the analysis would differ. Here, the record showed a contract spat escalated into a political blacklist. Allowing that precedent would let any administration punish companies for public criticism—hardly a recipe for secure, innovative supply chains.

Broader principle
Executive agencies have wide latitude in procurement and defense contracting. But “regarding” a U.S. company as a supply-chain risk cannot be a euphemism for retaliation without evidence, process, or statutory basis. Judicial injunctions enforce the rule of law, protect free speech and due process, and prevent abuse of power. That is exactly what the Framers designed the judiciary to do—even (especially) when the executive invokes national security. The injunction here was lawful, appropriate, and necessary.

[JesseAz (RIP CK)]

What law does grok think exists that allows for judges to undermine executive procurement?

[mad.casual]

Grok is not a lawyer. If you assume Grok is a lawyer, Claude is also a lawyer. If Grok and Claude are both lawyers, if you think the judge is right, Grok is both yours and a good lawyer. If you think the the judge is wrong, Grok is not your lawyer, and still better than Claude.

[JesseAz (RIP CK)]

To me it just shows the training data, sourced frome reddit, is always wrong.

The fact is LLM are conditioned to always agree with a) the user or b) the primary narratives. Making them utterly useless.

[mad.casual]

Except Claude neither agreed with me, the primary narrative, nor even answered the question. If anything, it raised a factual incorrectness that makes an out for its own (legal) demise.

To be clear: I'm solidly in the "You can have your money from Uncle Sugar, or you can have your ethics, but you can't have both; and China doesn't care about your ethics, they're just going to build a killer robot." camp.

If you want to build an ethical AI, your largest customer shouldn't be the DOW. Anthropic got in when they thought they were going to be spreading LGBTQIA+ rights to the Taliban in Afghanistan. That's their fault. There may be a minor procedural issue where the DOW has to give 30 or 90 days notice or pay out on existing contracts for 6 mos. whether it uses the tools or not, but the DOW absolutely has the right to forbid its employees from using any given toolchain at any given moment.

[JesseAz (RIP CK)]

A) with a check against B)

[Rick James]

Meet the philosopher teaching Claude to be "good". Trigger warning, she uses the word "like" a lot. A lot... a .... lot...

[Rick James]

[cut previous response where Claude assures me that the majority belief does not define morality]

My followup:

If majority does not define morality, that what was moral 100 years ago, but not moral today, is no guide as to whether the things we consider moral actually are.

2:25 PM
That's a sharp observation — you've identified a real tension. Let me steelman and then stress-test it.

What you're pointing to

If morality isn't just "what most people believe," then the fact that moral consensus shifts over time doesn't automatically mean we're getting better. We could just be getting differently wrong. Our current confident moral views might look as mistaken to people 100 years from now as Victorian views look to us. That's a genuinely humbling point.

The stronger version of your argument

Moral progress narratives assume a direction — that abolishing slavery was an improvement, not just a change. But if majority belief doesn't define morality, what standard are we using to call it progress? We're seemingly using our current values to judge past ones, which is circular.

Where it gets complicated though

There's a difference between:

Moral fashion changing (what's considered polite, what's taboo)
Moral reasoning improving
Many philosophers argue we've developed better tools — more consistency, wider circles of consideration, better empirical knowledge about suffering and human nature. The abolition of slavery wasn't just a shift in taste; it followed from applying principles of human dignity more consistently than before.

The honest conclusion

You're right that shifting consensus gives us no guarantee our current views are correct. But it doesn't follow that all moral views are equally valid, or that reasoning carefully about ethics is pointless. It does follow that:

Humility about current moral certainties is warranted
Confident moral dismissal of minority or traditional views deserves scrutiny
The grounds for a moral belief matter more than its popularity
The uncertainty cuts in all directions — including against confidently declaring traditional views wrong.

[Longtobefree]

Get back to me after the appeals.