AI Regulation Tracker / Courts and litigation
Third Circuit Rejects 'AI Decided My Case' Claim
In Watson v. Reese, a non-precedential per curiam opinion filed June 24, 2026, the U.S. Court of Appeals for the Third Circuit turned away a pro se litigant's motion arguing that a district court ruling was "tainted by artificial intelligence." Disagreeing with a decision, the court said, does not make it fraudulent. This opinion is not binding precedent.
On June 24, 2026, the U.S. Court of Appeals for the Third Circuit filed a short per curiam opinion in Watson v. Reese, No. 25-3118, an appeal from the Western District of Pennsylvania. The panel was Judges Bibas, Chung, and Bove. The case was submitted on June 22 and decided two days later. Before I go further, one thing has to be clear, because the court itself made it clear: this opinion is expressly non-precedential. It says on its face that it is "not an opinion of the full Court and pursuant to I.O.P. 5.7 does not constitute binding precedent." So nothing here creates new law. It is a data point, not a rule.
What the litigant argued
The appellant was proceeding pro se, meaning he represented himself without a lawyer. After losing in the district court, he filed a motion to vacate the judgment for what is called "fraud upon the court." That is a serious and narrow claim. It is meant for the rare situation where the integrity of the judicial process itself has been corrupted, not for a party who simply thinks the judge got it wrong.
His theory was novel. He contended that the district court's opinion against him was "tainted by artificial intelligence" and was "technically nonexistent." In plain terms, he was arguing that because he believed a machine had written or influenced the ruling, the ruling was a fraud on the court and should be wiped out.
What the court held
The Third Circuit did not spend much time on it. In a footnote, the court denied the motion and explained why in a single, clean sentence. It wrote that it was denying the motion to vacate, "which baldly contends that his case was decided by artificial intelligence," and then added: "Watson may disagree with the District Court's analysis, but that does not render it 'fraudulent.'"
That is the whole of it, and it is worth reading slowly. The court did not rule on whether AI can ever be involved in judicial decision-making. It did not make findings about how the opinion below was written. It simply held that a bare, unsupported allegation that a ruling was AI-generated is not fraud upon the court. Losing and then labeling the loss "artificial intelligence" does not clear the very high bar that a fraud-upon-the-court claim requires. Disagreement is not fraud.
I want to keep the claim ceiling honest here. This is not a sanctions ruling. Nobody was punished. The court did not announce a test for future cases, and because the opinion is non-precedential, it does not bind even other Third Circuit panels. Read it for what it is: a federal appeals court declining, briefly, to entertain an "the AI did it" attack on a judge's work.
Why this one is different from the cases we usually cover
This tracker has spent a lot of time on the other side of this coin. The running story in AI and litigation has been lawyers and self-represented parties getting caught submitting briefs with fake, AI-hallucinated case citations, and then facing sanctions, fee awards, and referrals to disciplinary bodies. In those cases, AI is the source of the misconduct, and the court is the one policing it.
Watson v. Reese flips the direction. Here the litigant is not the one who leaned on AI. He is the one accusing the court of using it, and using that accusation as a lever to try to undo a judgment he did not like. It is AI weaponization pointed at the bench rather than coming from the party. As far as the accusations the tracker follows go, this is a new vector.
What makes it useful is not the outcome, which is unsurprising, but the fact that a federal appellate court has now had to say out loud, in writing, that "this opinion was written by AI" is not by itself a reason to vacate anything. That is a marker for the next litigant who is tempted to make the same move.
What this means for lawyers and courts
For litigators, the practical read is short. An unsupported claim that a ruling was AI-generated is not a path to relief. Fraud upon the court is a demanding standard that has always required real evidence of corruption of the judicial process, and a hunch about how an opinion sounds does not come close. If you are representing a client who is convinced the court used AI, the honest counsel is that suspicion is not proof, and that dressing a disagreement up as fraud is more likely to hurt your credibility than help your case.
For courts, the case is a small preview of a problem that is going to grow. As AI tools spread into every corner of legal work, "the AI did it" is becoming a two-way accusation. Judges already have to watch for AI-fabricated citations coming in from parties. Now they also have to be ready for parties who accuse the court of the same thing on the way out. The Third Circuit's answer, that disagreement is not fraud, is the sensible starting point, and it may be cited as a non-precedential data point even though it does not have to be.
For everyone else, including US finance and compliance professionals watching how AI accountability plays out in practice, the lesson travels. Blaming the machine is going to be a reflex in both directions, from the person who used AI and from the person who suspects it was used against them. In court, and in any serious proceeding, the burden still lands on the person making the claim to actually show something. A label is not evidence.
What to do now
Read the opinion, not the headline. The holding is narrow and non-precedential: a bald allegation that a case was decided by AI is not fraud upon the court. It is not a ruling that judges cannot use AI, and it is not a ruling that they do. If you are litigating and suspect AI influenced a decision, gather actual evidence before you file anything, because a fraud-upon-the-court motion built on suspicion will fail and may cost you standing with the bench. And if you are advising clients or drafting your own filings, remember that the same scrutiny courts apply to AI-fabricated citations is exactly the scrutiny they will apply to your accusations. Show your work in both directions.
Questions professionals are asking
Did the Third Circuit rule that courts cannot use AI?
No. The court made no finding about whether or how AI was used in the decision below. It held only that a bare, unsupported allegation that a ruling was AI-generated does not amount to fraud upon the court. The opinion is narrow and does not address the broader question of AI in judicial decision-making.
Is this opinion binding precedent?
No. The opinion states on its face that it is "not an opinion of the full Court and pursuant to I.O.P. 5.7 does not constitute binding precedent." It is non-precedential. It can be cited as persuasive, but it does not control other cases, even within the Third Circuit.
What exactly did the litigant claim?
The appellant, who represented himself, moved to vacate the district court's judgment for "fraud upon the court." He argued that the opinion against him was "tainted by artificial intelligence" and "technically nonexistent." The court denied the motion, noting he may disagree with the analysis but that disagreement does not make a ruling fraudulent.
How is this different from the AI fake-citation cases?
Those cases involve parties or lawyers submitting AI-hallucinated citations and facing sanctions. This case is the reverse: a litigant accused the court of using AI and tried to use that accusation to vacate a judgment. It is AI weaponization aimed at the bench rather than misconduct by the filer, which makes it a new counterpoint to the fabricated-citation line.
Were sanctions imposed?
No. This was not a sanctions ruling. The court simply denied the motion to vacate in a footnote. No party was penalized, and the court did not announce a new legal test.
RELATED BRIEFINGS
Browse the full AI Regulation News tracker
Informational analysis for working professionals, not legal advice. Confirm how any decision or standard applies to your situation with qualified professionals in the relevant jurisdiction.