California Judge Slaps Lawyer with $10,000 Fine for ChatGPT-Generated Brief

A California appeals court has fined Los Angeles attorney Amir Mostafavi $10,000 after he submitted a legal brief riddled with fake case citations produced by ChatGPT.

The 2nd District Court of Appeal ruled that 21 of the 23 citations in Mostafavi’s opening brief were fabricated. The judges called the filing “frivolous” and a waste of both the court’s time and taxpayer resources.

According to CalMatters, the sanction appears to be the largest of its kind in California against an attorney for misuse of AI-generated text.

In its opinion — issued 10 days ago and certified for publication — the court underscored that attorneys are personally responsible for verifying every citation they submit. The panel stressed that filings containing unverified or fabricated material will not be tolerated.

The ruling serves as a pointed warning to the legal community: AI tools may assist in drafting, but reliance on unverified machine-generated content can carry steep professional and financial consequences.

In May, a US district court judge in California ordered two law firms to pay $31,100 in fees over “bogus AI-generated research,” saying they nearly cited fabricated material in an order and that “strong deterrence is needed.”

Mostafavi used ChatGPT to ‘improve’ appeal draft

Mostafavi told the court he drafted the appeal on his own, but later ran it through ChatGPT in hopes of improving the writing. He admitted he did not review the AI-generated output before filing, and said he was unaware that the tool might insert case citations or fabricate material.

Speaking to CalMatters, Mostafavi argued that it is unrealistic to expect lawyers to abandon AI altogether, likening its adoption to the shift from law libraries to online databases. Still, he cautioned that until the technology can reliably avoid generating false information, attorneys should use it carefully.

“I’m paying the price,” Mostafavi said, warning that others could fall into the same trap.

Experts warn of surge in AI fabrications

Damien Charlotin, a researcher who tracks global cases of fabricated AI citations, told CalMatters that the number of court filings seeded with false case law has surged — from just a few per month to several each day. He noted that AI models are particularly prone to hallucination when asked to bolster complex legal arguments, increasing the risk that briefs will contain invented material.

Legal scholars share that concern. Mark McKenna of UCLA described reliance on ChatGPT as “an abdication of your responsibility as a party representing someone.” Meanwhile, Andrew Selbst, also a law professor, cautioned that the technology is being “shoved down all our throats” in law schools and law firms without serious reflection on its consequences.

Too high a price for errors

AI systems are notorious for producing false information, with independent tests showing high rates of hallucination even on simple, easily verifiable tasks. The pattern underscores a persistent reality: despite their growing use, generative models cannot yet guarantee accuracy — and the costs of those errors can be steep.

“In the meantime we’re going to have some victims, we’re going to have some damages, we’re going to have some wreckages,” Mostafavi said after his sanction.

But when justice is at stake, waiting for AI to mature is not an option. Accepting “some victims, some damages, and some wreckages” as the price of technological progress is simply unacceptable.