Why AI Cites Really Bother the Courts

Want to know why bad AI cites really bother the courts? Jeff and Tim discuss two recent fake-AI-cites cases imposing sanctions and State Bar referrals, and draw this conclusion: It’s not that AI is bad at law—in one of these cases, the court noted that none of the AI mistakes even went in the direction of helping the offending party. Rather, the problem is that AI is just bad at citing and quoting the law. And the courts are super-protective against our legal canon becoming polluted with hallucinations.

• Blame game backfires: In Shayan v. Shakib, appellant’s counsel blamed non-attorney staff for adding the bad AI cites to the brief. The mistakes didn’t really change the arguments, and the court ordered counsel to file a corrected version. But the outcome is going to be the same, plus $7500 sanctions and a State Bar referral.
• Gatekeeping function: Courts emphasize that even when fabricated citations don't advance a party's position, they still threaten "the integrity of courts and the legal profession" by risking that fake law becomes cited as real precedent.
• We discuss updates in the Boies Schiller/Scientology case, and whether these recent cases predict the result.
• Voluntary dismissal dilemma: Tim’s firm filed an amicus brief in the Maniago case, arguing that voluntary dismissals with prejudice should be treated as appealable final judgments, challenging the rule that clerk-entered dismissals are merely "ministerial acts."
• Heated bench: A Texas redistricting case features an unusually scathing dissent beginning with "The main winners from Judge Brown's opinion are George Soros and Gavin Newsom," raising questions about appropriate judicial temperament.