Two Lawyers. Six Months. Zero Excuses.

The Ninth Circuit Court of Appeals has suspended attorneys Mike Singh Sethi and William Rounds for six months after they filed briefs loaded with fabricated case citations, misattributed quotations, and gross misrepresentations of real legal decisions.

The panel — Judges Richard Paez, Carlos Bea, and Danielle Forrest — issued a lengthy disciplinary opinion warning every attorney practicing before the court.

The message: clean up your AI problem, or face consequences.

They Lied About It Twice

When confronted, Sethi and Rounds claimed the errors were innocent typographical mistakes. Then they denied that generative AI had anything to do with it.

The court wasn't buying it. According to the opinion, published by Eugene Volokh at the Volokh Conspiracy via Reason, the court had already identified other cases where Sethi or Rounds filed briefs with similar problems. That pattern made the "typo" story collapse.

Denying AI involvement when the court can see a pattern across multiple filings isn't a defense. It's an aggravating factor.

The Two Flavors of AI Failure

The Ninth Circuit's opinion breaks down AI hallucinations into two categories. Both are dangerous. One is more dangerous than most people realize.

Fabrications are the obvious ones — the AI invents a case citation that doesn't exist at all. Courts can catch these with a basic search. Embarrassing, but detectable.

Inaccuracies are the silent killer. The AI cites a real case but describes it wrong — claims it stands for a legal proposition it doesn't actually support. According to the court's opinion, these "may prove more dangerous to our profession in the long run" because they don't fail a facial check. You have to actually read the source to catch them.

The court cited a 2024 study by Magesh et al., published in the Journal of Empirical Legal Studies, titled Hallucination-Free? Assessing the Reliability of Leading AI Legal Research Tools. The numbers are striking: Westlaw's AI hallucinated 17% of answers. LexisNexis's AI hallucinated 33% of answers. These are the premium legal research tools attorneys pay top dollar for.

One in three answers from Lexis AI is wrong or fabricated.

This Isn't New — It's Getting Worse

Stanford Law School's RegLab Director Professor Daniel Ho and JD/PhD researcher Mirac Suzgun have been studying this problem for years. In a May 2025 Stanford Legal podcast episode, Ho said: "The idea — that I think is actually right — is not that AI is going to replace lawyers. It's that lawyers who use AI, and know how to work effectively with it, will replace lawyers who don't."

That framing is correct. The problem isn't AI. The problem is lawyers who treat AI like a research assistant that doesn't need supervision.

Ho and Suzgun co-authored Profiling Legal Hallucinations in Large Language Models, published in the Oxford Journal of Legal Analysis in 2024, which examines this same failure mode.

What Mainstream Coverage Is Getting Wrong

Most coverage of this story focuses on the "AI gone wrong" angle — as if the technology is the villain.

The technology didn't lie to the court. Sethi and Rounds lied to the court. The AI produced garbage output. The attorneys had a professional obligation to verify every citation before signing their name to it. They didn't. Then they covered it up.

Framing this as an AI story lets the lawyers off the hook. It's a professional responsibility story with AI as the instrument.

Also underreported: the Ninth Circuit's specific concern about inaccuracies over fabrications. Every news account on this topic focuses on "fake cases" — the fabrications. But the court explicitly warned that subtle inaccuracies in real citations may be the more serious long-term threat. A fake case gets caught. A real case that's been mischaracterized by AI might sail through unchallenged.

The Bar Needs to Act Like It

The Ninth Circuit has now put the entire federal appellate bar on notice. Disclose AI hallucinations fast when you find them. Read every single citation — yes, every one — whether AI drafted it or not. Overreliance on generative AI is NOT a defense.

State bars should be watching this ruling closely. The disciplinary standards here apply in federal court. But attorneys practicing in state courts face the same AI tools and the same failure modes with, in many jurisdictions, zero formal guidance on the books.

Six months is a significant suspension. It will cost Sethi and Rounds clients, income, and reputation. But for the profession overall, that punishment only matters if other attorneys actually change their behavior.

With Westlaw hallucinating 17% of answers and Lexis at 33%, similar problems are almost certainly occurring across multiple jurisdictions and courtrooms.