Case Brief Update · 09.05.26
AI in the courtroom.
A living, source-linked reading list addressing two questions: what happens when lawyers misuse generative AI, and when prompts, outputs, and AI-assisted work become discoverable.
What this list covers
Reported rulings, not every docket mention.
These briefs collect the leading public decisions verified through September 5, 2026 that directly address AI-assisted legal work or the discovery and protection of prompts and outputs. This editorial list excludes pro se filings, pending show-cause matters, suspected AI use, and lawsuits about AI products.
Browse the broader incident trackerAttorney AI use · 27 briefs
Lawyers remain responsible for every filed word.
The decisions state the same rule: using AI is not itself misconduct. Filing false material, failing to supervise, skipping primary-source review, or giving the court an evasive explanation is misconduct.
Lisandrillo v. Palozzi
Florida Fourth District Court of Appeal
The court referred the matter to the Florida Bar after counsel defended filings containing a false quotation and arguments contradicted by the record. Counsel’s claim of independent verification did not resolve the defects in the filings.
Read case briefTRAX International v. United States
U.S. Court of Federal Claims
The court ordered government counsel to explain factual misrepresentations about the Army’s AI-generated proposal evaluations and the omission of those evaluations from the administrative record.
Read case briefIn re Stephen D. Aarons
New Mexico Supreme Court
The New Mexico Supreme Court held appellate counsel in direct contempt after a ChatGPT-assisted brief included fabricated witnesses, false testimony, and misrepresented legal authority.
Read case briefDouglas v. Deutsche Bank
District of Columbia Court of Appeals
The court struck an entire appellee brief after counsel used Google’s generative-AI search and cited four nonexistent cases without verification.
Read case briefQuinteros v. Harbor Distributing
California Supreme Court / California Court of Appeal, First District, Division Two
The California Supreme Court denied review and depublication, leaving in place a published decision affirming sanctions for an AI-assisted brief with fabricated authority.
Read case briefTOV Realty v. Suarez
Connecticut Supreme Court
ChatGPT corrupted previously verified citations during a stylistic edit, and the firm’s reviewers did not check them again.
Read case briefDel Biaggio v. Bansen
California Court of Appeal, First District, Division Four
The court sanctioned appellate counsel for fabricated quotations in an AI-assisted brief, rejecting his plan to rely on a paralegal to verify the authorities.
Read case briefLandberg v. City of New York
New York Supreme Court, Appellate Division, Second Department
An appellate brief contained fictitious cases and fabricated quotations generated during AI-assisted research, followed by an initial denial and a later admission at argument.
Read case briefLNU v. Blanche
U.S. Court of Appeals for the Ninth Circuit
Immigration briefs contained nonexistent cases, misattributed quotations, and major misstatements, followed by repeated denials or minimization of AI use.
Read case briefPayne v. State
Supreme Court of Georgia
Unverified AI citations migrated from a prosecutor’s papers into a proposed judicial order and then into the appeal.
Read case briefUnited States v. Farris
U.S. Court of Appeals for the Sixth Circuit
Briefs prepared with Westlaw CoCounsel contained fabricated quotations and misstated holdings that remained unchecked after editing.
Read case briefMcCarthy v. DEA
U.S. Court of Appeals for the Third Circuit
Counsel relied on inaccurate AI summaries supplied by a nonlawyer and minimized the problem after the government flagged it.
Read case briefFletcher v. Experian
U.S. Court of Appeals for the Fifth Circuit
An AI-assisted reply included 16 fabricated quotations and five serious misrepresentations, followed by a misleading show-cause response.
Read case briefMatter of Zareh
New York Supreme Court, Appellate Division, First Department
The court disciplined a supervising lawyer for inadequate review, inadequate supervision, and a bad-faith defense of a defective filing.
Read case briefDeutsche Bank v. LeTennier
New York Supreme Court, Appellate Division, Third Department
At least 23 fabricated cases appeared across five filings, continued after warning, and were followed by a reluctant AI admission.
Read case briefIn re Sanctions Order of Kenney
Louisiana Court of Appeal, Fifth Circuit
The lawyer used one AI tool to check another; the appellate court held a fictitious citation was prima facie evidence of an objectively unreasonable filing.
Read case briefMattox v. Product Innovations
U.S. District Court for the Eastern District of Oklahoma
Fourteen nonexistent and 14 erroneous or misquoted authorities spread through 11 pleadings before the court imposed layered sanctions.
Read case briefNoland v. Land of the Free
California Court of Appeal, Second District
Twenty-one of 23 quotations in an opening brief were fabricated, with more errors in reply and an admission that counsel had not read the cases.
Read case briefJohnson v. Dunn
U.S. District Court for the Northern District of Alabama
Five fabricated or misrepresented authorities in discovery motions led to public reprimands, disqualification, notice obligations, and bar referrals.
Read case briefIn re Baby Boy & A.H.
Illinois Appellate Court, Fourth District
Court-appointed counsel cited eight nonexistent cases and then inaccurately claimed that some of them were valid.
Read case briefGarner v. Kadince
Utah Court of Appeals
An unlicensed clerk used ChatGPT to invent precedent, the signer failed to verify it, and the firm had no AI policy.
Read case briefDehghani v. Castro
U.S. District Court for the District of New Mexico
Counsel bought a freelancer’s draft, never read the cited cases, and filed six fake authorities plus 13 that did not support the propositions stated.
Read case briefLacey v. State Farm
U.S. District Court for the Central District of California
An AI-assisted discovery brief passed between two firms with fake or misquoted authority; even the revised filing remained defective.
Read case briefWadsworth v. Walmart
U.S. District Court for the District of Wyoming
A lawyer used an apparent in-house AI tool and filed motions in limine in which eight of nine cited decisions did not exist.
Read case briefPark v. Kim
U.S. Court of Appeals for the Second Circuit
An appellate reply cited a nonexistent ChatGPT case, leading to a published reminder that Rule 11’s verification duty applies to AI output.
Read case briefPeople v. Crabill
Colorado Office of the Presiding Disciplinary Judge
Counsel filed a ChatGPT-drafted motion without reading its authorities, delayed correction, and falsely blamed an intern.
Read case briefMata v. Avianca
U.S. District Court for the Southern District of New York
The landmark sanctions decision involved invented cases and quotations, followed by submission of fabricated “opinions” when the court asked questions.
Read case briefPrompts, privilege & discovery · 18 briefs
Prompts qualify as evidence, work product, or both.
Protection turns on purpose, direction, confidentiality, and use. Courts have protected private litigation testing and ordered disclosure of chats created independently, shared selectively, used by a witness, or embedded in expert methodology.
Villanueva v. LVMPD
U.S. District Court, District of Nevada
The court required disclosure of an expert’s ChatGPT-generated case summaries but protected his prompts, deposition preparation, and draft-report material.
Read case briefCifuentes v. Sid Tool
U.S. District Court, Middle District of Florida
The court denied discovery requests concerning a self-represented plaintiff’s alleged AI use because the defendant failed to adequately support them. It separately reminded the plaintiff that Rule 11 applies to AI-assisted filings.
Read case briefShealy v. Seaside Investments
Massachusetts Superior Court, Suffolk BLS2
A client’s partner independently used ChatGPT to draft litigation material. The resulting prompts and drafts were not protected work product.
Read case briefAssini v. Hayward
New York Supreme Court, Nassau County
The court quashed subpoenas seeking a pro se litigant’s OpenAI prompts, uploads, outputs, drafts, and account material.
Read case briefTate Group v. Legacy Automotive
Texas Business Court, Eleventh Division
After reviewing ChatGPT conversations in camera, the court protected most of them but ordered selected pages and upload information produced.
Read case briefConservation Law Foundation v. Shell
U.S. District Court for the District of Connecticut
A magistrate judge treated an expert team’s GPT-4o prompts as discoverable methodology, but the district judge stayed the order for review.
Read case briefJones v. Delta Air Lines
U.S. District Court for the Eastern District of Michigan
After a pro se deponent admitted having ChatGPT open, the court barred AI or outside assistance while answering deposition questions.
Read case briefTym v. Cerno
U.S. District Court for the District of New Mexico
The court accepted, for that case, that AI interactions used to prepare filings or hearings do not automatically lose work-product protection.
Read case briefMorgan v. V2X
U.S. District Court for the District of Colorado
The court treated a pro se litigant’s AI-assisted preparation as work product but required him to identify the AI tool.
Read case briefJefferies v. Harcros Chemicals
U.S. District Court for the District of Kansas
The court restricted uploads of discovery to public or open AI tools while allowing secure closed systems with safeguards.
Read case briefUnited States v. Heppner
U.S. District Court for the Southern District of New York
A represented defendant’s independent Claude sessions were neither attorney-client privileged nor attorney work product.
Read case briefWarner v. Gilbarco
U.S. District Court for the Eastern District of Michigan
The court refused a request for a pro se plaintiff’s AI tools, prompts, drafts, and outputs and treated litigation-preparation material as work product.
Read case briefIn re OpenAI (user logs)
U.S. District Court for the Southern District of New York
The district judge affirmed an order requiring protected production of 20 million de-identified consumer ChatGPT logs for sampling.
Read case briefConcord Music v. Anthropic
U.S. District Court for the Northern District of California
A three-order sequence protected attorney testing, allowed aggregate facts, and compelled post-suit prompts used by planned witnesses.
Read case briefR (Munir) v. Secretary of State
United Kingdom Upper Tribunal (Immigration and Asylum Chamber)
The tribunal warned that uploading client documents to open-source ChatGPT could breach confidentiality and waive privilege.
Read case briefIn re OpenAI (ChatExplorer)
U.S. District Court for the Southern District of New York
The court denied a request for all prompts and outputs from The New York Times’s internal infringement-testing tool as irrelevant and disproportionate.
Read case briefT.B. v. Big Brothers Big Sisters
New York Supreme Court, New York County
Selective use of part of a ChatGPT conversation waived work product and led to production of the entire transcript.
Read case briefTremblay v. OpenAI
U.S. District Court for the Northern District of California
Attorney-crafted prompts and negative tests were opinion work product; only the interactions underlying examples used in the complaint had to be produced.
Read case briefReading note
Discovery orders turn on specific facts.
Several trial-level rulings remain under review. Each brief states the material limits and links to the underlying order. This collection is educational, not legal advice, and will change as courts revisit these questions.