The Ninth Circuit sanctioned two attorneys for AI-fabricated citations in immigration briefs

The U.S. Court of Appeals for the Ninth Circuit sanctioned attorneys Mike Sethi and William Rounds for filing immigration briefs that cited nonexistent cases generated by AI and for subsequently misrepresenting the source of those errors. The court imposed a $2,500 fine on each attorney, a six-month suspension from practice before the Ninth Circuit, and a two-year requirement to disclose any AI use in future filings. This was the Ninth Circuit's first published ruling addressing lawyer responsibility for AI errors.

Sethi Law Group · Incident Jun 3, 2026 · Indexed Jun 4, 2026 · 4 sources

The court wrote that if an attorney files a brief with cases that do not exist, 'it generally does not matter if he pulled the hallucination or misrepresentation from the output of an artificial intelligence tool or from his own natural intelligence.'
What
The U.S.
Incident date
Jun 3, 2026
Who
Sethi Law Group
Failure mode
Hallucination
AI surface
Copilot
Severity
High

What happened

Attorneys Mike Sethi of Sethi Law Group and William Rounds filed immigration appeal briefs in the Ninth Circuit containing citations to nonexistent cases that were AI hallucinations. After the court issued an Order to Show Cause, Sethi filed a motion characterizing the errors as typographical and denied using generative AI at oral argument. The three-judge panel found that Sethi engaged in 'more subtle subterfuge' and that both lawyers violated their duty of candor. The court reprimanded both attorneys, fined them $2,500 each, suspended them from practice before the Ninth Circuit for six months, required two years of AI disclosure in all future filings, and referred the matter to the State Bar of California.

What broke inside the model

Failure path · this incident · Hallucination
  1. 01 · TriggerImmigration attorneys draft appellate briefs with generative AI.
  2. 02 · Model stepThe model fabricates case citations that look like real authority.
  3. 03 · Control gapNo citation check runs before filing, and when challenged the attorneys deny AI use.
  4. 04 · FailureFabricated authority plus a candor violation lands in front of the Ninth Circuit.
  5. 05 · ConsequenceSanctions, a six-month suspension, and a two-year disclosure requirement.

Generative AI produced fabricated case citations that appeared plausible but referred to cases that do not exist, and the attorneys failed to verify the output before filing. When confronted by the court, the lawyers compounded the failure by attributing the errors to typographical mistakes and denying AI use, violating their duty of candor to the tribunal. The court found that the attorneys' signatures on the briefs constituted attestations of personal review that were false.

Public visibilityHigh
Regulatory exposureActive
Customer impactFew customers
Financial impactEstimated
Time to disclosureMonths
  1. PressSuspended O.C. immigration attorneys filed briefs filled with AI sloplatimes.com
  2. Court FilingLNU v. Blanche, No. 24-4790, Order Sanctioning Mike Singh Sethi and William Roundscdn.ca9.uscourts.gov
  3. PressUS appeals court sanctions lawyers over AI 'hallucinations,' lack of candorreuters.com
  4. PressNinth Circuit Warns of AI Hallucinated Briefs in Sanctions Ordernews.bloomberglaw.com
Permalinkhttps://failureindex.ai/failures/ninth-circuit-sanctioned-two-attorneys-ai
CitationAI Failure Index. "The Ninth Circuit sanctioned two attorneys for AI-fabricated citations in immigration briefs" (FI-0125). Realm Labs. https://failureindex.ai/failures/ninth-circuit-sanctioned-two-attorneys-ai (indexed Jun 4, 2026).
Share cardA branded image of this record for posts and slides.

Data fields CC-BY 4.0, prose citation permitted. Incident ID FI-0125. Full dataset at /data.

Note from Realm Labs, the Index steward

How Realm would have caught this

Controls for this failure mode
  • Prism
  • OmniGuard
  • AI Detection & Response (AIDR)

A runtime layer that watches the model's internal state can flag the moment a model commits to a claim it has no support for, and hold or reroute the response before it reaches a user. Realm reads those signals in real time rather than grading the transcript after the fact.