Part of the AI Regulation News hub.
The Idaho Supreme Court has stated that using generative AI does not excuse inaccurate briefing by lawyers or self-represented litigants, confirmed its power under I.A.R. 11.2 to sanction fabricated authority, and imposed no sanction in the appeal before it
The Court never found that AI was used and said the origin of the errors mattered less than their cost. The appellant still lost nine of her ten issues, and the Court was careful to say those losses were not sanctions.
Bottom line: Binding decision in this appeal. The Court affirmed a $1,176,000 defamation judgment, stated that AI use does not excuse inaccurate briefing, confirmed it can sanction under I.A.R. 11.2, and chose to impose no sanction.
Who this affects: Idaho appellate lawyers and the firms that supervise their brief writing, self-represented appellants, legal aid and court self-help staff, and professional responsibility counsel advising on AI drafting policies.
Issue date: Opinion filed 22 September 2026, Lewiston, April 2026 Term. Chief Justice Bevan wrote; four justices concurred.
What changed: Idaho's highest court put in writing that reliance on generative AI is no excuse for inaccurate briefing, whether by attorneys or self-represented litigants, and tied the duty to the signature certification in I.A.R. 11.2.
Analysis: The test the Court applied is verification, not the tool. It declined to decide whether AI was used and rested on the finding that the cited authorities were not meaningfully checked before filing.
Primary sources: Posey v. Bushnell, Idaho Supreme Court Docket No. 52072, opinion (PDF)
- Instrument (EN)
- Opinion, Eric Posey v. Summer Bushnell
- Authority
- Supreme Court of the State of Idaho
- Jurisdiction
- Idaho, United States. Appeal from the District Court of the First Judicial District, Kootenai County
- Status
- Opinion filed. Judgment of the district court affirmed; costs awarded to the respondent under I.A.R. 40(a)
- Bindingness
- Decision of Idaho's highest court, binding on the parties. No sanction was imposed under I.A.R. 11.2
- Issue date / next deadline
- Filed 22 September 2026. No deadline is set by the opinion
- Document
- Docket No. 52072, 17 pages. Opinion by Chief Justice Bevan; Justices Brody, Moeller and Zahn and Justice Pro Tem Monson concurred
- Legal basis
- Idaho Appellate Rule 11.2 (signature certification and sanctions); I.A.R. 35(a)(6) on argument and authority
- Primary source
- https://api.isc.idaho.gov/uploads/Documents/ISC/Civil/2026/1790088181964-52072.pdf
What did the Court say about generative AI?
The statement comes in the opinion's introduction, before any facts. The Court said it was taking the opportunity "to address the risks associated with reliance on generative artificial intelligence" and "to make clear that such reliance is fraught with potential peril", and that AI use "does not excuse inaccuracies in briefing, whether it be used by attorneys or self-represented litigants."
The substantive discussion sits in Part IV.A.4, headed "Misrepresented facts and authorities can merit sanction". There the Court said the nature and extent of the errors "suggest that she may have relied on generative artificial intelligence", citing Kruse v. Karlen, a 2024 Missouri Court of Appeals decision, for the proposition that such tools often hallucinate.
It went no further than suggest. The appellant told the Court she had not used GenAI. The Court said review of her briefing and the record "raise serious doubt as to that assertion", then set the question aside: the origin of the deficiencies, it wrote, "is less concerning than the burdens they impose on this Court."
What was wrong with the briefing
The Court counted. Of twenty-two purported quotations attributed to caselaw, statutes, court rules, pattern jury instructions and the Idaho Constitution, it found only six accurate, and those with minor spelling errors.
It gave examples. A cited "Jones v. Lynn, 155 Idaho 439, 445 (2013)" does not exist; the only Idaho Jones v. Lynn is a 2021 summary judgment appeal that does not discuss jury selection. Two further cases, McAdam v. McAdam and Dairyland Insurance Co. v. Hawkins, could not be found. The appellant said McAdam was a North Dakota case she had failed to identify, and the Court said the record did not support that explanation.
Separately, a footnote records that her briefs repeatedly cited Idaho Code section 16-1604, a Child Protective Act provision, when discussing punitive damages. The Court assumed she meant section 6-1604.
The Court's view of why this matters is short and quotable: "fabricated authority is more troubling than the absence of authority because it introduces false information into the judicial process."
The sanction power the Court confirmed, and did not use
After identifying the errors, the Court ordered supplemental briefing on why the appeal should not be dismissed as a sanction, citing the Ninth Circuit's 2024 decision in Grant v. City of Long Beach. The appellant apologised in writing and at oral argument and said the errors were inadvertent. The Court accepted those representations.
It then located the duty in I.A.R. 11.2. Every party signs its briefs and thereby certifies that, "to the best of the signer's knowledge, information, and belief after reasonable inquiry[,]" the filing is "well grounded in fact and is warranted by existing law". The Court said all litigants, represented or not, carry that obligation, and that it "possesses ample authority to sanction a litigant" for briefing with pervasive inaccuracies of this kind.
It declined to use that authority. The Court noted that the respondent had asked for no attorney fees, said that omission did not limit its power to impose a monetary sanction, and gave its reasons for holding back: nine issues already disposed of for briefing failures, and an existing judgment of $1,176,000 plus post-judgment interest. It concluded that the conduct "could warrant sanctions under Rule 11.2" but imposed none.
Why losing nine issues was not a sanction
This is the part practitioners are most likely to misread. The respondent asked for the appeal to be dismissed outright. The Court refused, but declined to consider issues one through nine because they lacked adequate argument and authority, rested on misstated facts in the record, or were not preserved below.
The Court was explicit that "those dismissals are not sanctions under Idaho Appellate Rule 11.2". It described them as substantial consequences flowing from the failure to meet briefing requirements. On our reading, the practical effect for a litigant who files unverified citations is the same whether or not a sanction is labelled as one: arguments built on fabricated authority are treated as unsupported and are not reached.
The one surviving issue, whether the plaintiff was a limited-purpose public figure, failed on harmless error. The jury had already found actual malice by clear and convincing evidence when it awarded punitive damages, so a different public-figure ruling would not have changed the outcome.
What the opinion does not say
The opinion does not describe itself as the Court's first treatment of generative AI in briefing, and we do not describe it that way. It announces no new rule of appellate procedure and does not amend I.A.R. 11.2; it applies the existing certification requirement.
It makes no finding that AI was in fact used, it imposes no sanction, and it sets no disclosure requirement for AI-assisted filings. Readers should treat the AI passages as the Court's stated position applied to these facts.
What we did not verify
What we opened: the 17 page opinion PDF from the Idaho Supreme Court's own server, retrieved 24 September 2026 UTC and read end to end. Its SHA-256 matches the copy taken by the sweep.
What we did not open: the appellant's opening, reply and supplemental briefs; the Court's order requiring supplemental briefing; the district court record; and the cited decisions in Kruse v. Karlen and Grant v. City of Long Beach. We did not check whether a petition for rehearing has been filed.
What we refuse to claim: that this is the Idaho Supreme Court's first opinion on generative AI, because the opinion does not say so; that the appellant used AI, because the Court did not find that she did; that any sanction was imposed; and that the nine issues were dismissed as a sanction, because the Court said they were not. Quotations are reproduced with ASCII punctuation in place of the opinion's typographic apostrophes and quotation marks.
Informational analysis for working professionals, not legal advice. Confirm how any rule applies to your situation with qualified counsel.
The Idaho Supreme Court placed the risk on the signature, not the software. A brief that cites authority nobody checked is treated as unsupported whether or not AI wrote it, and the Court has said in terms that it can sanction under I.A.R. 11.2 even where the other side asks for nothing. Verification before filing is the control that the opinion rewards.
Source File
https://api.isc.idaho.gov/uploads/Documents/ISC/Civil/2026/1790088181964-52072.pdf
Open the opinion and read three passages: the last paragraph of the introduction on page 2, the twenty-two quotation count in Part IV.A.1 on page 7, and Part IV.A.4 on pages 13 and 14, where the Court discusses generative AI, I.A.R. 11.2 and its decision to impose no sanction.
We also take this opportunity to address the risks associated with reliance on generative artificial intelligence and to make clear that such reliance is fraught with potential peril. As explained below, the use of artificial intelligence does not excuse inaccuracies in briefing, whether it be used by attorneys or self-represented litigants. ยท Posey v. Bushnell, Idaho Supreme Court Docket No. 52072, introduction, p. 2, 22 September 2026
FAQ
Did the Idaho Supreme Court sanction the appellant for AI-generated citations?
No. The Court said her conduct could warrant sanctions under I.A.R. 11.2 and that it had ample authority to impose them, but it exercised its discretion to impose no sanction in this appeal.
Did the Court find that generative AI was used?
No. It said the errors suggested she may have relied on generative AI and that her denial raised serious doubt, but it treated the origin of the errors as less important than the burden they imposed.
Does the rule apply to self-represented litigants?
Yes. The Court said all litigants, whether represented by counsel or proceeding pro se, are responsible for the accuracy of what they present, and that AI use does not excuse inaccuracies by attorneys or self-represented litigants.
What happened to the arguments that relied on fabricated cases?
The Court declined to consider them for lack of adequate authority. It disposed of nine of her ten issues for briefing and preservation failures and said those dismissals were not sanctions under I.A.R. 11.2.
Related briefings
Sponsored Training
Practical AI training for regulated professionals, built around verification, documentation and a defensible process. See the courses.