Over 1,200 documented cases. That's how many times courts have now caught lawyers filing AI-generated content with fabricated citations, according to tracking by Damien Charlotin at HEC Paris business school. Roughly 800 of those are in U.S. courts alone.
The pace is accelerating. Charlotin recently identified 10 cases from 10 different courts on a single day. And the penalties are getting steeper.
A $109,700 Record
A federal court in Oregon set what may be a new high-water mark last month, ordering a lawyer to pay $109,700 in sanctions and costs for filing AI-generated errors. That dwarfs the $3,000-per-attorney fines handed to lawyers for MyPillow CEO Mike Lindell, which was one of the earliest high-profile AI citation cases.
The problem has reached state supreme courts. In February, Nebraska's high court grilled attorney Greg Lake about a brief containing citations to cases that don't exist. The justices weren't persuaded by his explanation and referred him for professional discipline. A similar scene played out in the Georgia Supreme Court in March.
The Core Problem: Hallucinated Case Law
Here's what keeps happening. A lawyer asks ChatGPT or a similar tool to find relevant case law. The AI generates plausible-sounding case names, court names, and citations. The lawyer copies them into a brief without checking whether those cases actually exist. They don't. The opposing counsel or judge runs a search, finds nothing, and sanctions follow.
This isn't an AI problem. It's a due diligence problem. As Carla Wale at the University of Washington Law Library put it, lawyers are required under professional conduct rules to read the cases they cite and verify their accuracy. That obligation existed long before ChatGPT.
Joe Patrice of Above the Law noted that AI is genuinely useful for tasks like combing through large volumes of evidence or case law. The issue is lawyers treating AI output as a finished product rather than a starting point that requires verification.
Courts Are Adding New Rules
Some courts have started requiring lawyers to label anything produced with AI assistance, specifying which parts are human-generated versus AI-generated. This disclosure requirement adds friction but creates accountability.
The pattern here is clear. AI tools can speed up legal research, but they can't replace the fundamental step of verifying that the cases you're citing are real. Lawyers who skip that step aren't saving time. They're generating six-figure sanctions, disciplinary referrals, and the kind of professional embarrassment that follows a career.