A federal court ruled judicial immunity applies even if a judge let AI write her ruling

In an order filed August 12, 2026, U.S. District Judge Gloria M. Navarro (D. Nev.) dismissed a lawsuit against Nevada judge Mari Parlade, adopting a magistrate's recommendation that judicial immunity barred the case with prejudice. The dismissal took the plaintiff's core allegation at face value: that Parlade "relied wholly on artificial intelligence to issue a judicial ruling, without any discretionary human thought."

What the court did and didn't decide

The order never determines whether Parlade used AI to produce her ruling, because the answer doesn't change the outcome. Under the four-factor test courts use to decide whether conduct is judicial in nature, the order reasons that producing a ruling in a pending case is "a normal judicial function," that the dispute centered on the plaintiff's own case before that judge, and that nothing alleged took the conduct outside her official capacity. Once conduct clears that bar, judicial immunity is absolute. It covers even action that is erroneous, malicious, or in excess of authority, so the court dismissed the case without ruling on the AI-delegation claim.

A procedural bar

Volokh's post reproduces the order's language and draws the same line the ruling does: even if the allegations are correct and the judge relied entirely on AI, "she can't be sued for that in federal court." He notes the plaintiff still has recourse through state appellate review, mandamus-style petitions, and judicial-conduct disciplinary proceedings. What's closed is a federal damages suit against the judge herself.

The ruling neither blesses AI-authored rulings nor confirms one happened. It holds that one accountability mechanism, suing the decision-maker directly, doesn't reach this category of decision-maker however the decision was made. That's a narrower result than "judges can use AI." It also previews a harder problem for any organization: as AI takes on more of the reasoning behind consequential decisions, the mechanisms for checking that reasoning don't update on their own while the output still looks like a normal decision.