A federal court in Nevada just ruled that a state judge accused of relying entirely on AI to write a ruling is still protected by judicial immunity. The lawsuit was dismissed on August 17, 2026, and the decision has nothing to do with whether AI belongs in a judge's chambers — it turns on a much older doctrine that predates AI by decades.
Short answer: In Phillips v. Parlade, Judge Gloria Navarro (D. Nev.) dismissed a suit alleging a state judge relied wholly on AI to issue a ruling, because judicial immunity covers any act that is judicial in nature, regardless of how the ruling was produced. The plaintiff's remaining options are state appellate review, a mandamus petition, or a judicial-conduct complaint, not a federal lawsuit against the judge.
I spend most weeks testing what AI chat tools do when you push them toward tasks they shouldn't fully own, and legal drafting is one of the clearest examples. In my testing, when I asked ChatGPT and Claude to draft a court order from scratch, both appended a line saying the output needed review by a licensed attorney before it could be filed or relied on. That gap between what the tools flag on their own and what apparently happened here is exactly why this ruling matters.
What happened in Phillips v. Parlade
A litigant sued a state court judge in federal court, arguing that judicial immunity shouldn't apply because the judge allegedly relied wholly on AI to produce a judicial order, with no independent human discretion behind it. Judge Navarro didn't rule on whether that allegation was true. She ruled that it didn't matter for immunity purposes, and that's the core of the AI judicial immunity ruling everyone's now talking about.
Judicial immunity is one of the oldest doctrines in American law: judges are absolutely immune from civil liability for judicial acts, "even if their action was in error, done maliciously, or in excess of their authority," as the court put it. There are only two recognized exceptions — a "clear absence of all jurisdiction," or conduct that isn't judicial in nature at all (an off-the-bench act, for example). Issuing a ruling in a pending case is a textbook judicial act. The court found nothing in the complaint suggesting the judge acted outside her jurisdiction or outside her role, so immunity applied and the case was dismissed.
Why "how it was written" didn't change the outcome
The doctrine protects the function of ruling on a case, not the process behind it. A judge who writes an order longhand, dictates it to a clerk, copies a boilerplate template, or allegedly leans on an AI tool is still performing the same judicial function under this framework. The court's reasoning treats the AI-reliance allegation the same way it would treat an allegation that a judge ruled while distracted, careless, or wrong on the law: those are grounds for appeal, not grounds to sue the judge personally.
What remedies are actually still available
Dismissal on immunity grounds doesn't mean an AI-drafted error is unreviewable, it means the review has to happen through the right channel. The Reason/Volokh writeup of the AI judicial immunity ruling (Aug. 17, 2026) points to three that remain open: state appellate review of the underlying order, a mandamus petition asking a higher court to compel correction, and a complaint to the state's judicial disciplinary body. None of those require proving the judge acted with malice or outside her authority, which is the high bar a civil suit against a judge has to clear.
This isn't an isolated incident
This ruling landed against a backdrop of real, documented AI drafting failures inside federal courts. In July 2025, U.S. District Judge Henry T. Wingate (S.D. Miss.) issued a temporary restraining order that named the wrong parties, misquoted state law, and referenced people who weren't part of the case; he later replaced it with a backdated "corrected" version. Days later, U.S. District Judge Julien Xavier Neals (D.N.J.) withdrew a ruling that misquoted a defendant and misstated which motions had been granted or denied. Senate Judiciary Chairman Chuck Grassley opened a formal inquiry into both judges that October, asking directly whether they or their staff used generative AI and demanding the original, error-filled orders be re-docketed for transparency.
Prompts you can copy to sanity-check an AI-assisted legal draft
These are checks worth running on anything AI helped draft, order, brief, or contract, before anyone signs off on it.
- Fabrication check: "Go through this document line by line. For every case citation, statute, or factual claim, tell me whether you generated it from training knowledge or from a source I gave you, and flag anything you're not certain is accurate."
- Party and fact check: "List every named party, date, and dollar figure in this document. Cross-check each one against [the source document/case file] and flag any mismatch."
- Human-review gate: "Rewrite this so it's clear it's a draft requiring attorney or judicial review before filing, not a finished, ready-to-issue document."
Common mistakes to avoid
The mistake I see most often, in comment threads reacting to this ruling, is treating "judicial immunity" as if it means courts have blessed AI-written rulings as fine. It doesn't. The court never decided whether the judge actually used AI, or whether doing so wholesale would be misconduct; it only decided that a federal civil suit is the wrong venue to litigate that question. Second is assuming immunity applies to everyone in this chain — it protects the judge's judicial acts specifically, not a clerk, vendor, or court IT system that mishandled a filing outside a judicial role. Third, for anyone using AI to draft anything with legal weight, contracts, filings, internal policies, don't mistake a clean-looking draft for a verified one. Wingate's and Neals's withdrawn orders both read fluently; the errors were in the facts, not the prose, which is exactly the kind of mistake a quick skim won't catch.
Where an AI-drafted judicial error can actually be challenged
| Path | What it addresses | Requires proving judge acted with malice/outside authority? | Outcome in this case |
|---|---|---|---|
| Federal civil suit against the judge | Personal liability for the ruling | Yes (or a jurisdiction/non-judicial-act exception) | Dismissed — barred by judicial immunity |
| State appellate review | Whether the ruling itself was legally correct | No | Still available |
| Mandamus petition | Compelling a higher court to fix or vacate the order | No | Still available |
| State judicial disciplinary complaint | Whether the judge's conduct violated ethics rules | No | Still available |
Tools that make this easier
If you're using AI anywhere in a document workflow, legal or otherwise, the same verification habits apply. My how to spot AI writing guide covers the tells that show up when AI-generated text gets rushed into a final document without a real edit pass, the same fluent-but-wrong pattern that tripped up Judge Wingate's order. If you're drafting with a general assistant, my guides to how to use ChatGPT and how to use Claude AI both cover prompting it to flag uncertain claims instead of stating them as fact. I've also covered a related case where someone hid a prompt injection in a legal filing aimed at any AI reviewing the document, another sign that courts are now a real attack surface for AI misuse, not just a place AI might quietly make mistakes. Before trusting any AI tool with document drafting or review, my AI tool ratings hub scores tools on actual tested behavior rather than marketing claims, and my free AI tools roundup is a reasonable place to start if budget is the only thing stopping you from adding a second tool as a cross-check.
My take
The headline reads scarier than the AI judicial immunity ruling actually is. Judicial immunity didn't get expanded here, it got applied the way it always has been: to the judicial act itself, not the tool used to produce it. What should worry people more is the pattern sitting underneath this case, two federal judges have already had to publicly withdraw AI-tainted rulings in the past year, and a Senate committee is asking pointed questions about it. The lesson isn't "you can't do anything about a bad AI-assisted ruling." It's that the fix runs through appeal, mandamus, or a disciplinary complaint, not a lawsuit against the judge personally, and that anyone using AI to draft something with legal consequences should build in the same fact-by-fact verification pass these courts clearly skipped.
Frequently Asked Questions
Can you sue a judge for using AI to write a ruling?
Not directly, based on this ruling. Judicial immunity protects judges from civil liability for judicial acts even if the act was flawed, and a federal court found that issuing a ruling is a judicial act regardless of whether AI was involved in producing it.
What is Phillips v. Parlade about?
It's a federal case in the District of Nevada where a litigant sued a state court judge, alleging she relied wholly on AI to issue a judicial order without independent human discretion. Judge Gloria Navarro dismissed the suit on August 17, 2026, ruling that judicial immunity applied regardless of the allegation.
If a lawsuit against the judge is barred, how can an AI-related judicial error be fixed?
Through state appellate review of the ruling, a mandamus petition asking a higher court to compel a correction, or a complaint to the state's judicial disciplinary body — none of which require proving the judge acted maliciously or outside her authority.
Have other judges actually gotten in trouble for AI-drafted errors?
Yes. In 2025, U.S. District Judges Henry T. Wingate and Julien Xavier Neals both had to withdraw or replace rulings containing AI-pattern errors, like misquoted law and misstated case outcomes, which drew a formal inquiry from the Senate Judiciary Committee.
Does this ruling mean courts have approved judges using AI to write orders?
No. The court didn't decide whether the judge used AI or whether doing so would be misconduct. It only decided that a federal civil suit against the judge personally is the wrong way to challenge that conduct.