A Nevada litigant thought he had found the crack in judicial immunity. In a federal suit against family court judge Mari Parladé, a plaintiff named Phillips argued that she never really decided his case at all. An AI did, he claimed. And if a machine wrote the ruling, his logic went, no judicial act ever took place, so the centuries-old shield that protects judges from being sued should not apply.
Last week a federal court told him no, in terms broader than he probably expected. U.S. District Judge Gloria Navarro dismissed the case and wrote that judicial immunity would hold even if a judge relied “wholly on artificial intelligence” to issue a ruling. Producing a decision in a pending case sits squarely within a judge’s ordinary duties, Navarro found, and immunity attaches to the act itself, not to the quality of the reasoning behind it. The order began circulating through legal blogs on August 17 and has been rattling around the profession since.
The evidence was thin. The principle was not.
Phillips built his AI theory on remarkably little. Judge Parladé read her ruling from a prepared script. She stumbled over a word. At one point she said “appearance of any impartiality” where the phrase should have been “appearance of impropriety.” That was the case: a verbal slip and a script, the kind of thing any judge reading from notes might produce on an ordinary afternoon in court.
Navarro never had to decide whether the allegation was true. Her order makes no finding that Parladé used AI, no finding that she delegated anything, and no ruling that judges may hand their judgment to software. The point was narrower and, for exactly that reason, more sweeping: even taking the claim at face value, immunity applies. Phillips also tried the one recognized escape hatch, arguing that total delegation to a machine placed the ruling outside the judge’s jurisdiction, a rare exception that can defeat immunity. Navarro rejected that too. Issuing orders in a pending case is precisely what judges have jurisdiction to do, however they arrive at them.
Lawyers get sanctioned. Judges are immune.
The asymmetry is hard to miss. Attorneys who let chatbots write their briefs have been fined and referred to bar regulators across the country, and courts show less patience with each new incident. State Farm’s lawyers were caught citing cases that never existed just this month. A lawyer who files hallucinated citations answers for it personally, sometimes in front of a very unhappy judge. A judge who allegedly signs an AI-drafted order cannot be sued at all.
There is a reason the doctrine works this way. Judicial immunity exists so that every losing party cannot relitigate a case by suing the judge who decided it, and the remedy for a bad ruling is supposed to be appeal, not damages. The doctrine has survived accusations of bias, corruption and plain incompetence over more than a century of American case law. It was always going to survive an accusation of ChatGPT.
The accountability question immunity does not answer
Immunity from lawsuits is not the same as permission. Judges remain subject to appellate review and to judicial conduct commissions, and the AI question has already reached both. In late 2025, two federal judges acknowledged that AI-assisted drafting in their chambers had produced orders containing errors, after congressional scrutiny pushed the details into the open. The federal judiciary has since issued interim guidance cautioning judges against delegating core decision-making to AI tools.
That is where this fight will actually play out. A litigant who believes a machine decided their custody dispute or sentencing has no damages claim, but they can appeal, and they can file a complaint with the body that disciplines judges. Neither route has yet produced a clear national standard on when, whether and how judges must disclose AI use in their chambers. Nothing requires a judge to tell the parties at all.
Navarro’s order settles only the narrowest question: suing the judge is not an option. Everything else, from disclosure rules to whether litigants deserve to know when software helped write their fate, remains open. Expect the conduct commissions and the appellate courts, not damages suits, to answer it first.
For more coverage of AI and the law, visit Mylistingo.
