The Order

On June 3, 2026, the U.S. Court of Appeals for the Ninth Circuit published a disciplinary order in LNU v. Blanche, an immigration appeal handled by attorneys Mike Singh Sethi and William Rounds of Sethi Law Group / US Legal Group, APC.

The underlying immigration case involved a petition for review from the Board of Immigration Appeals. That part of the case was serious, high-stakes work. People were challenging immigration decisions, not arguing over a parking ticket or a contract typo. The disciplinary order came later, after the court identified a different problem in the lawyers' filings: fake cases, fake quotations, and descriptions of real cases that did not match the actual opinions.

The Ninth Circuit said Sethi's opening brief cited two cases that never existed: Eduardo v. Garland and Lay v. Holder. The brief also attributed quotations to real Ninth Circuit decisions where the quoted language did not appear. When the court pressed the issue, the lawyers initially tried to characterize some of the errors as typographical mistakes. The court was not buying it.

This was not a situation where a citation had a transposed digit or a reporter abbreviation went sideways. The court described nonexistent authorities, misattributed quotations, and gross misrepresentations of real cases. Calling that a typo is the legal equivalent of calling a kitchen fire "aggressive toasting."

What the Court Was Sanctioning

One important distinction runs through the order: the court did not punish the lawyers simply because generative AI was involved. The panel was explicit that attorneys may write personally, delegate to subordinates, or use technology, including generative AI, so long as the final filing complies with the rules.

That distinction matters because courts are not trying to ban every drafting tool that produces a bad first draft. The legal system has always had bad first drafts. The sanctionable conduct was filing material that contained fabricated authority, failing to verify the authorities before submitting the brief, and then failing to deal candidly with the court once the defects surfaced.

In practical terms, the rule is painfully simple: if a brief cites a case, a lawyer needs to confirm the case exists and says what the brief claims it says. If the brief quotes an opinion, a lawyer needs to read the opinion and confirm the words are there. This is not advanced litigation craft. This is table stakes. It is the professional version of checking whether the ladder is leaning against a wall before climbing it.

Generative AI made the failure easier to create at scale. A language model can produce a plausible-looking appellate brief with case names, reporter citations, and judicial-sounding language. Some of those outputs may be correct. Some may be wrong. Some may be fiction wearing a necktie. The tool does not know the difference in the way a legal database does. The attorney has to.

Candor Was the Multiplier

The Ninth Circuit's discipline turned heavily on candor. The court explained that if the lawyers had disclosed the AI issue promptly, acknowledged the failure, and apologized for not checking the brief, lighter sanctions might have been appropriate. Instead, the panel found repeated failures to be straightforward about where the errors came from.

That is where these cases keep getting worse for lawyers. The first failure is trusting generated text without verification. The second failure is treating the court like it cannot tell the difference between an ordinary citation typo and hallucinated authority. Judges read for a living. Opposing counsel reads for a living. The odds that nobody checks a suspicious case citation in a federal appeal are not zero, but they are not odds anyone sensible should build a career around.

The Los Angeles Times coverage noted that the attorneys were fined and suspended, and that the disciplinary ruling did not undo the earlier immigration decision. That nuance is worth keeping. This story is not about punishing the clients for their lawyers' drafting workflow. It is about the court protecting the integrity of filings submitted to it.

The Discipline

The court imposed several concrete penalties.

Each attorney was personally sanctioned $2,500. Both were suspended from practicing before the Ninth Circuit for six months. They were ordered to provide the disciplinary order to clients, opposing counsel, and presiding judges in their pending cases, and to provide it to every attorney in the firm. The court also directed the clerk to send the order to state bar and licensing authorities for any further proceedings those bodies consider appropriate.

The most unusual requirement may be the two-year disclosure rule. For two years, Sethi, Rounds, and all attorneys at their firm must include a sworn statement in future filings addressing whether generative AI was used, identifying the tool if one was used, and certifying that the signing attorney personally reviewed the filing and verified that every citation and quotation refers to real authority.

That is a professional scarlet letter, but one drafted in civil procedure language. Every future filing from the firm before the Ninth Circuit now carries an extra reminder that the court no longer trusts the ordinary verification process there.

Why This Fits the Graveyard

This is a clean Vibe Graveyard entry because the failure path is documented and the consequences are formal. A professional workflow relied on generative AI or AI-generated material, fabricated legal authority reached a court filing, and the court imposed sanctions, suspension, disclosure obligations, and potential licensing follow-up.

The broader pattern is also depressingly familiar. Early AI citation cases were treated as embarrassing one-offs. By 2026, federal appellate courts are writing disciplinary orders with the tone of people who have run out of patience for the same preventable mistake. Nobody is asking lawyers to understand transformer architecture. The requirement is much smaller and much harder to excuse: read the cases you cite.