The Ninth Circuit AI Sanctions Ruling: What Every Filing Now Requires

On June 3, 2026, the Ninth Circuit did something the AI-and-law story had been missing. In a published, precedential opinion, it sanctioned two California attorneys for fabricated citations, and it explained in writing why. Until now, the warnings came from individual district judges in one-off orders that bound no one beyond that courtroom. This is a federal appellate court, on the record, setting a rule the circuit must follow. That changes the stakes for anyone who lets AI near a brief.

Here is the part the headlines keep getting wrong. The court did not punish the lawyers for using AI. Read the opinion and the holding is narrower and far more useful than the panic suggests: the sanction was for the lack of candor that followed the mistake, not for the tool that caused it. The lawyers who got hurt were not the ones who used AI. They were the ones who would not own what it did. This briefing covers what the ruling held, why precedential matters, and the pre-filing workflow that turns this from a threat into an advantage.

Key takeaways

On June 3, 2026, the Ninth Circuit issued a published, precedential opinion sanctioning two attorneys for AI-fabricated citations. As binding circuit precedent, it governs every court in the circuit, not just one judge's courtroom.
The court punished the cover-up, not the AI. The sanction rested on a lack of candor about the fabricated authority, not on the decision to use an AI tool to draft.
The penalties were real: a monetary sanction each, a six-month suspension, and a two-year mandatory AI-disclosure obligation on the attorneys' future filings. Disclosure was imposed as a remedy for hiding the problem, not as a rule against using AI.
A four-step pre-filing verification protocol, cite-check, quote-check, holding-check, and a candor log, satisfies your Rule 11 and FRAP duties and makes a tool-fluent attorney safer than a junior who hides the source.

What the ruling held, and why precedential matters

The case, decided June 3, 2026, involved two California attorneys whose brief cited authority that did not exist. The Ninth Circuit panel imposed a monetary sanction on each, suspended them for six months, and ordered that for two years they disclose any use of AI in their filings. Those are serious consequences for a citation error. The reason they landed so hard has less to do with the fabrication itself than with what the lawyers did after it surfaced.

The word that matters most in the coverage is precedential. Most AI-sanction stories so far have been district-court standing orders or individual sanction rulings: a single judge, reacting to a single brief, in a way that binds only the parties in front of that judge. A published Ninth Circuit opinion is different in kind. It is binding precedent across the entire circuit, the nine western states the court covers, and it gives every district judge inside that circuit a clear, citable rule to apply the next time a fabricated citation appears. It also signals to other circuits how a federal appellate court is thinking about this. That is why this ruling is the one to internalize, even if you have read a dozen sanction headlines already.

If you want the underlying mechanics, why confident-sounding fake cites appear in the first place and how to catch them, our briefing on AI citation hallucinations in legal filings covers that ground. This page is about what the court did with one, and what its reasoning asks of you.

The cover-up, not the AI: the distinction that changes everything

This is the heart of it, and it is the opposite of the lesson most lawyers are taking away. The Ninth Circuit did not hold that using AI is sanctionable. It held that filing fabricated authority and then failing to be straight about it is. The mandatory disclosure was a remedy aimed at conduct, a response to attorneys who were not candid, not a blanket signal that AI use must always be announced.

Sit with what that means for how you practice. A model produced a wrong answer. Models do that, and everyone, including the court, knows it. The fatal step was not the bad output. It was a licensed attorney signing it, filing it, and then declining to own the error cleanly when it came to light. The court's logic rewards exactly the behavior senior lawyers already value: candor with the tribunal. An experienced attorney who verifies the work, and who, if something slips through, corrects it forthrightly, is in a stronger position after this ruling than before. The lawyer exposed by this precedent is the one who treats AI as a secret to hide rather than a tool to supervise.

So the takeaway is not fear of the tool. It is this: own your filings. The signature is yours, the verification is yours, and if a mistake surfaces, the candor is yours too. Get those right and the ruling is on your side.

The pre-filing AI-verification workflow

Here is the fix, and it is the reason a tool-fluent attorney has nothing to fear from this ruling. The court asked for verified accuracy and candor. The sanctioned lawyers delivered neither. Both point to the same small discipline: a structured check that runs before anything reaches the clerk. Build it once and it costs you minutes per filing.

Step one: cite-check every authority

List every case, statute, rule, and regulation the filing cites. Not a sample, every one. AI tools fabricate selectively, so a spot-check is exactly how a fake citation survives. If your draft passed through a model, assume nothing about which cites are real until you have looked at each one.

Step two: quote-check in a primary source

Open each authority in a source you trust: the official reporter, the court's own database, a paid research service. Confirm the case exists, that the citation resolves to it, and that the quote you attribute to it appears verbatim. A citation that does not resolve in a real database is the single clearest red flag, and it is the one this ruling now makes costly.

Step three: holding-check the proposition

Existence and an accurate quote are still not enough. Read the passage and confirm the case actually stands for the point you are using it to make. AI models will pair a real case with a holding it never reached. This step is where your legal judgment does the work no tool can do, and it is the work the duty of candor is really asking you to sign for.

Step four: keep a candor log

Note in the file who checked the citations and when, and flag anything AI-assisted that you confirmed. This is the step the Ninth Circuit's reasoning rewards directly. If a problem ever surfaces, a contemporaneous record turns your defense from a hope into a documented fact, and it makes correcting course forthright instead of awkward. Candor is easiest when you have already written down what you did.

If you want this built into a repeatable senior-review system rather than a checklist you might forget under deadline, that is exactly what we teach in The Leveraged Attorney: how to let AI draft while your verification, your judgment, and your candor stay firmly in your hands.

Rule 11 and competence: the duties underneath the ruling

The Ninth Circuit did not invent a new obligation. It applied old ones to a new tool. Federal Rule of Civil Procedure 11 and the parallel appellate rules already require that the contentions in your filings have a basis and that you have made a reasonable inquiry. The Federal Rules of Appellate Procedure carry the same expectation into the circuit courts. Your duty of competence has always meant understanding the technology you use well enough to use it responsibly, and your duty of candor has always forbidden presenting authority you have not confirmed.

What this ruling adds is consequence and clarity. It says, with the force of circuit precedent, that signing an AI-assisted filing without verifying it is a Rule 11 problem, and that hiding the lapse afterward is a candor problem that courts will treat as the more serious of the two. If you already verify your citations and deal straight with the court, the ruling changes little except that the line is now drawn in published ink. For a fuller treatment of where these duties bite across your practice, see our AI competence duty checklist for lawyers.

What to do now

If you practice in the Ninth Circuit, the precedent is binding today, so the move is immediate: add the four-step protocol to your filing process this week and make sure anyone who drafts under your name knows the verification and the candor log are not optional. If you practice elsewhere, adopt the same workflow now, because a published federal appellate opinion is precisely the kind of authority other circuits and state courts look to when they write their own rules.

Then set the standard for your team. The signature on the brief carries the responsibility, which means a partner who signs owns an associate's unverified cites. A short written protocol, every cited authority pulled and confirmed before filing plus a note of who checked, protects the people working under you as much as it protects you. Make AI part of the workflow openly, make verification the non-negotiable last step, and make candor the reflex if anything ever slips. For how state and federal courts have been handling this more broadly, our companion piece on AI court rules across New York and Florida maps the wider landscape this ruling now sits atop.

What AI does not replace

It is worth being plain about the boundary. AI can draft a brief, summarize a record, and surface authorities faster than any associate. It cannot make a reasonable inquiry under Rule 11. It cannot read a case and tell you, with a lawyer's judgment, whether it truly supports your argument or merely sounds like it might. And it cannot be candid with a court on your behalf, because candor is a choice only the signer can make. The Ninth Circuit drew that line in precedent: the verification and the honesty are the parts of the job that are yours alone. They are not what AI took from you. They are, more than ever, the value you bring.

Frequently asked questions

Did the Ninth Circuit ban lawyers from using AI?

No. The June 3, 2026 ruling did not prohibit AI. The court sanctioned the attorneys for filing fabricated citations and for the lack of candor about them, not for using an AI tool. The two-year disclosure obligation was a remedy aimed at that conduct, not a rule against AI use. You can keep using AI to draft; you must verify before you file and be straight with the court.

Why does it matter that the ruling is precedential?

Earlier AI-sanction warnings came from individual district judges in standing orders that bound only their own courtrooms. A published, precedential Ninth Circuit opinion is binding across the entire circuit and gives every court in it a citable rule to apply. It also signals to other circuits and state courts how a federal appellate court is approaching AI-fabricated filings.

What workflow protects me under this ruling?

Run a four-step pre-filing check on every filing: cite-check every authority listed, quote-check each one in a primary source, holding-check that the case supports your point, and keep a candor log of who verified what and when. It takes minutes, satisfies your Rule 11 and FRAP duties, and gives you a documented record if a question ever arises.

Build the verification discipline into how you work

The attorneys who thrive with AI are not the fastest drafters. They are the ones whose verification is so reliable that speed never costs them accuracy, and whose candor with the court is automatic. That discipline is teachable, and it is the spine of how we train senior lawyers to use these tools.

Learn the AI workflow built for attorneys: The Leveraged Attorney Join The Leverage Club for $49 and get the templates and protocols Not sure where to start? Take the 2-minute course finder

Sources: ABA Journal, "Attorneys' lack of candor over AI errors leads to stricter sanctions" (June 2026); Ninth Circuit published opinion, No. 24-4790 (June 3, 2026), cdn.ca9.uscourts.gov; Reason / Volokh Conspiracy, "Ninth Circuit on AI Hallucinations" (June 3, 2026).