Legal information, not legal advice · Every entry verified to its primary source · Independent of any bar association
The Legal AI Compliance Register

legalaicompliance.help · est. 2026 · honest labels, primary sources, public changelog

Changelog

Every change to the tracker is recorded here with a date: new instruments, corrections, and re-verifications. Freshness you can audit, not just a "last updated" badge. Subscribe by RSS or email alert to get changes the week they ship.

2026-07-22: Court AI disclosure guide now indexes every adopted court rule on generative AI

The court AI disclosure orders explainer gained an at-a-glance index of the ten adopted court-rule instruments the tracker already carries, across eight states: New York's 22 NYCRR Part 161, the Illinois Supreme Court policy, Florida's Rule of General Practice and Judicial Administration 2.515(d)(2), Colorado's Rule Change 2026(02), Arkansas's Administrative Order 25 and the companion Rules of Professional Conduct amendments, the Montana Fourth and Thirteenth Judicial District local rules, the Oklahoma Court of Criminal Appeals Rule 1.17, and Rhode Island's Rule 1.1 comment and interim generative AI guidelines. Each entry states its effective date and its posture precisely: only the two Montana district rules impose an affirmative duty to disclose that AI was used, while the rest require certifying the accuracy of the filing. No new instruments or legal claims were added; every line restates data already verified in the tracker. The page title, description, and three FAQ entries were refreshed to match how lawyers search for this, and the year in the title moved to 2025-2026.

2026-07-21: Sanctions dataset grows from 6 to 19 cases: thirteen more AI and fabricated-citation sanctions verified against primary sources

A fetch pass verified thirteen new cases against the courts' own opinions and orders, taking the dataset from six to nineteen. Federal additions: Fletcher v. Experian (5th Cir., Feb. 18, 2026, $2,500, generative AI found); Wadsworth v. Walmart (D. Wyo., Feb. 24, 2025, $5,000, from an internal Morgan & Morgan platform the court expressly found was not ChatGPT); In re Kheir (Bankr. S.D. Tex., Nov. 4, 2025, opposing-party fees and costs); Johnson v. Dunn (N.D. Ala., July 23, 2025, public reprimand and disqualification of three Butler Snow attorneys, ChatGPT conceded and found); United States v. Hayes (E.D. Cal., Jan. 17, 2025, $1,500, no express AI finding); United States v. Farris (6th Cir., Apr. 3, 2026, forfeiture of Criminal Justice Act compensation and removal, Westlaw's CoCounsel named); and Davis v. Marion County (S.D. Ind., $2,000 imposed May 20, 2026). State additions: Kruse v. Karlen (Mo. App., Feb. 13, 2024, $10,000 against a self-represented appellant); In re: Sanctions Order of Kerry Kenney (La. App. 5th Cir., Oct. 23, 2025, $1,368 against attorney Connie P. Trieu, with ChatGPT, Copilot, and Gemini named); and four Oregon Court of Appeals decisions that built a per-infraction fee schedule for AI-fabricated authority — Ringo v. Colquhoun (Dec. 3, 2025, $2,000), Powell v. Employment Department (Feb. 11, 2026, $500 against a self-represented petitioner), Doiban v. OLCC (Mar. 18, 2026, $10,000, the largest of the group), and Williams v. Honl (Apr. 22, 2026, $8,044.25). Every row records the court's AI-attribution exactly, from named tools to no express finding at all. Two citation cautions were resolved along the way: the reporter pincites some sources attach to Kheir and Kenney could not be verified in any primary or secondary record, so both are cited by docket; and the Oregon matter some trackers caption under the attorney name Watson is the same case as Ringo, and is listed once.

2026-07-19: Arkansas adopted both AI instruments on December 11, 2025: two entries flip from proposed to current

A live read of the Arkansas Judiciary's own opinions and rules databases resolved the adoption question this tracker had carried since the June 5, 2025 proposals went out for comment. The Arkansas Supreme Court adopted Administrative Order No. 25, Artificial Intelligence, by per curiam order on December 11, 2025 (2025 Ark. 220), effective immediately, after the comment period closed without substantive comments; the order now appears in the court's current Administrative Orders compilation with the history line Adopted December 11, 2025. The same day, the court adopted the companion amendments to the Arkansas Rules of Professional Conduct (2025 Ark. 221), adding Preamble paragraph 22, under which nothing in the rules prohibits an attorney's use of non-human assistance including AI but such use does not excuse or mitigate any violation, and retitling Rule 5.3 to Nonlawyer Assistance. Both dataset entries flip from proposed to current court rules with day-precision adoption dates, and the Arkansas state page has been rewritten to describe the adopted instruments, including an honest scope note: the administrative order's binding prohibition runs to court staff and court systems, while its awareness section reaches everyone entering client or court data into generative AI, attorneys included. One caution recorded: the court's posted current-ARPC compilation still showed rule current to June 12, 2025 at verification time, so the adoption per curiams are the operative primaries.

2026-07-19: Watching: California SB 574 would write attorney AI duties into statute

Added SB 574 (Umberg) as a proposed instrument after verifying the bill text and status on the Legislature's own site. As amended July 2, 2026, the bill would add Business and Professions Code section 6068.1, providing that an attorney shall not delegate the practice of law to generative artificial intelligence and imposing duties on attorneys who use it, including keeping confidential, personal identifying, and other nonpublic information out of public generative AI systems; amend Code of Civil Procedure section 128.7, the filing-certification and sanctions statute; and add sections 180 and 1282.1, requiring judicial officers to disclose generative AI reliance in rulings and making arbitrators responsible for all aspects of an award regardless of AI use. The bill passed the Senate and cleared the Assembly Privacy and Consumer Protection Committee 15 to 0 on July 1, 2026; it is in Assembly floor process and is not law. The California page now covers all three tracks: the approved practical guidance, the proposed conduct-rule amendments, and the bill.

2026-07-19: Fabrications watch adds a fifth item: a Colorado Bar AI opinion that does not exist

A competitor state-ethics page presents Colorado Bar Association Formal Ethics Opinion 145 as a March 2024 formal opinion specifically addressing AI use by attorneys, and builds billing, confidentiality, and disclosure guidance on it. The Colorado Bar's own formal-opinions index tells a different story: Formal Opinion 145 is Discrimination; Bias, dated May 14, 2022, and no AI opinion appears anywhere in the index, which runs through opinion 151. Colorado's actual AI instrument is the Supreme Court's Rule Change 2026(02), a conduct-rule amendment already in this dataset. The item follows this page's standing protocol: the subject page was re-fetched live on publication day, July 19, 2026, an independent Wayback Machine archive was captured the same day, and the claim is quoted verbatim against the primary record it fails. A second site that previously carried the same fabricated citation no longer does, and is therefore not documented; this page publishes only what is live.

2026-07-19: Sanctions dataset grows from 2 to 6 cases: Mata, Grant, Noland, and Prososki verified against primary documents

Four cases pass primary-source verification and enter the sanctions dataset. Mata v. Avianca, Inc., No. 22-cv-1461 (S.D.N.Y. June 22, 2023), the foundational case, verified against the court's Opinion and Order on Sanctions: a $5,000 penalty imposed jointly and severally on two attorneys and their firm under Rule 11, with mandatory notification letters to the client and to every judge falsely named as the author of a fake opinion; ChatGPT is expressly named. Grant v. City of Long Beach, Nos. 22-56121/22-56143 (9th Cir. Mar. 22, 2024), verified against the court's published opinion: opening brief stricken in its entirety and the appeal dismissed over fabricated and misrepresented case law; the opinion never mentions AI, and the dataset records that absence rather than assuming attribution. Noland v. Land of the Free, L.P., No. B331918 (Cal. Ct. App. Sept. 12, 2025), verified against the published opinion: $10,000 against counsel payable to the clerk, with a State Bar referral, after the court found 21 of 23 quotations in the opening brief fabricated by generative AI tools counsel conceded using. Prososki v. Regan, No. S-25-295 (Neb. Mar. 20, 2026), verified against the official Advance Sheets opinion: brief stricken, appeal dismissed, and counsel referred to the Counsel for Discipline; counsel denied AI use and the court held the analysis is the same regardless, so the dataset records no AI finding. Also verified this pass: United States v. Cohen, 18-CR-602 (S.D.N.Y. Mar. 20, 2024), in which the court expressly declined to impose sanctions for Google Bard-generated fake citations after finding no bad faith — recorded in our working roster and excluded from this dataset precisely because no sanction issued. The dataset lists only what the courts' own documents support: six cases, each with its AI-attribution status stated exactly.

2026-07-19: Four more state pages and a new formal opinion: coverage reaches 49 of 51 jurisdictions

Added deep-dive pages for Alabama and Kansas plus two no-instrument state pages for Connecticut and Indiana, all built from a live browser session against sources that block automated fetch. Alabama contributes a new instrument, and an upgrade on what was expected: the Alabama State Bar's numbered Formal Opinion 2026-01, Artificial Intelligence Use: Best Practices Under Existing Professional Conduct Rules, the newest of the bar's 141 formal opinions. Its premise is that AI does not create new ethical duties but recontextualizes competence, confidentiality, supervision, candor, fees, and client communication, and it closes with a per-matter checklist; the opinion document is undated, so the entry carries year precision, 2026, and nothing more specific. The Connecticut page records a dated verified negative, the CBA informal-ethics index read in full through opinion 25-04 with no AI opinion, alongside judiciary activity that is employee-facing by the Judicial Branch's own description. The Indiana page records a verified negative against an ISBA opinion index whose newest listed year is 2020. The Kansas page documents Supreme Court Administrative Order 2025-CM-017 of February 27, 2025, creating the court system's Ad Hoc Artificial Intelligence Committee, as judiciary activity, and states the bar-side posture precisely: the Kansas Bar Association does not publicly index its ethics advisory opinions, so no Kansas AI ethics opinion is publicly listed. Two jurisdictions remain deferred rather than guessed: Maryland's task-force report sits behind a member login, and South Dakota's judiciary guidance PDF is no longer reachable at its known address and was never archived, so nothing is asserted about either. The tracker now lists 55 instruments across 49 of 51 jurisdiction pages.

2026-07-01: Updated: California practical guidance approval date confirmed (May 14, 2026)

The State Bar of California's Ethics and Technology Resources page confirms that the Board of Trustees approved the updated revisions to the Practical Guidance for the Use of Generative Artificial Intelligence in the Practice of Law on May 14, 2026. Our California entry previously carried month-level precision (approval reported in May 2026); it now carries the verified day.

2026-07-01: New page: Corrections & Fabrications Watch

Launched a public log of legal-AI ethics citations published on other sites that fail primary-source verification, including a claimed Utah opinion that appears in no Utah State Bar index (the citation appears to be fabricated), a claimed Illinois State Bar Association generative-AI opinion that its published index does not contain (also apparently fabricated), a West Virginia no-guidance claim contradicted by the Lawyer Disciplinary Board's Legal Ethics Opinion 24-01 (June 2024), and three state mandates described on a vendor blog that we could not locate in any primary source. Every item carries a live re-fetch date, an independent Wayback Machine archive, and the primary record we checked it against. Corrections invited: verified fixes ship within a week.

2026-07-01: Added: Ninth Circuit advisory committee note on AI-produced filings

Added the Circuit Advisory Committee Note to Ninth Circuit Rule 32-1, adopted with the court's December 1, 2025 circuit-rule revisions. The note states that the rules do not directly regulate how a filing is produced, including by employing generative artificial intelligence, but that the signature required by FRAP 32(d) is an attestation that the signer has reviewed the filing and is responsible for its contents. The Ninth Circuit applied the note in Lnu v. Blanche, No. 24-4790 (June 3, 2026), a published order suspending two attorneys from practice before that court for six months and fining each $2,500 over fabricated citations and repeated lack of candor. Tracked as guidance; the tracker now lists 36 instruments.

2026-07-01: Watching: federal Rule 11 suggestion on citation accuracy (26-CV-21)

Added Rules Suggestion 26-CV-21, submitted May 29, 2026 by U.S. Magistrate Judge Patricia Barksdale and docketed with the Advisory Committee on Civil Rules, proposing a new FRCP 11(a) certification that cited legal authorities exist and are accurately cited, modeled on Florida's 2026 rule amendment. This is the earliest stage of federal rulemaking, a docketed suggestion pending consideration, not a proposed amendment published for public comment. Listed as proposed, not law.

2026-07-01: New dataset: AI sanctions case tracker (verified against courts' own documents)

Launched a new open dataset of court sanctions decisions from the generative-AI era, each verified against the court's own published document before listing, downloadable as JSON or CSV under CC BY 4.0 from the new sanctions page. It seeds with two 2026 federal appellate cases. Lnu v. Blanche, No. 24-4790 (9th Cir. June 3, 2026), a published disciplinary order, fined two attorneys $2,500 each and suspended them from practice before the Ninth Circuit for six months, with a two-year AI-use certification condition on future filings and a referral to the State Bar of California; no specific AI tool is named in the order, and the attorneys conceded unauthorized generative-AI use by unlicensed brief writers was more likely than not the source. Whiting v. City of Athens, Tenn., Nos. 24-5918/5919, 25-5424 (6th Cir. Mar. 13, 2026) imposed $15,000 in punitive sanctions on each of two attorneys plus appellees' appellate attorneys' fees and double costs over briefs with over two dozen fake citations and misrepresentations of fact; the court made no express AI finding, so the dataset records it as a fabricated-citations case, not a confirmed AI case. The dataset is deliberately small and grows only as cases pass primary-source verification.

2026-07-01: Seven new state deep-dive pages: coverage reaches 24 of 51 jurisdictions

Added full deep-dive pages for Alaska, Massachusetts, Michigan, Minnesota, Missouri, Utah, and Vermont, each written from that state's verified instruments in the dataset: Alaska Ethics Opinion 2025-1 (a formal opinion adopted April 23, 2025), the Massachusetts Board of Bar Overseers Office of Bar Counsel article The Wild West of Artificial Intelligence (2024), the State Bar of Michigan's Artificial Intelligence for Attorneys FAQs (November 18, 2024, fee questions added February 11, 2025), the MSBA AI Working Group Final Report (adopted summer 2024), Missouri Informal Opinion 2024-11 (April 25, 2024), the Utah State Bar's May 2023 ChatGPT article (with a pointer to the fabricated Utah opinion citation documented on the corrections page), and the Vermont judiciary's VJCAIC First Annual Report (March 2025). State deep-dive coverage now stands at 24 of the 51 U.S. jurisdictions, the 50 states plus the District of Columbia.

2026-07-01: Six more state pages and five instruments: coverage reaches 30 of 51 jurisdictions

Added deep-dive pages for Arkansas, Colorado, Georgia, and Hawaii, plus the tracker's first two no-instrument state pages, Delaware and Idaho. Five instruments enter the dataset: proposed Arkansas Supreme Court Administrative Order No. 25 on artificial intelligence (2025 Ark. 117) and the companion proposed amendments to the Arkansas Rules of Professional Conduct (2025 Ark. 115), both published for comment June 5, 2025 and both listed strictly as proposals pending confirmation of their final adoption status; Colorado Rule Change 2026(02), adopted January 8, 2026 and effective immediately, the first AI-specific amendment to a state's professional conduct rules adopted by a state supreme court that is recorded in this tracker; the State Bar of Georgia's Generative AI Toolkit (last updated February 20, 2026); and the Final Report of Hawaii's Committee on Artificial Intelligence and the Courts (December 15, 2025). The Delaware and Idaho pages document dated, verified negatives against the primary indexes we inspected, alongside each state's judiciary activity. State page coverage now stands at 30 of 51 jurisdictions, and the tracker lists 41 instruments.

2026-07-01: Seven more state pages and six instruments: coverage reaches 37 of 51 jurisdictions

Added deep-dive pages for Louisiana, Montana, and New Hampshire, plus four no-instrument state pages for Iowa, Maine, Nebraska, and Nevada. Six instruments enter the dataset. Louisiana's is the January 22, 2024 letter on the emergence of artificial intelligence from the Louisiana Supreme Court to the Louisiana State Bar Association, written by the Court's General Counsel at the Justices' direction and concluding that existing rules are robust and broad enough to cover AI in its current forms, tracked as guidance. Montana contributes the tracker's first district-level local court rules, each scoped to its own counties because Montana has no statewide instrument: Fourth Judicial District Rules of Practice Rule 3.G on pleading requirements for generative AI use (Missoula and Mineral Counties, October 31, 2025 rules revision) and Thirteenth Judicial District Local Rule 35 on artificial intelligence (Yellowstone County, adopted and effective January 1, 2026). New Hampshire contributes the NHBA Ethics Committee's three-article series, tracked as guidance rather than formal opinions: Ethics of Using Artificial Intelligence in Practice (May 15, 2024), Ethics of Drafting Documents with Artificial Intelligence (September 10, 2024), and Understanding Generative Artificial Intelligence and The Potential for Discrimination and Bias (October 30, 2024). The Iowa, Maine, Nebraska, and Nevada pages document dated, verified negatives against the indexes we inspected, alongside each state's judiciary, committee, legislative, and case activity. State page coverage now stands at 37 of 51 jurisdictions, and the tracker lists 47 instruments.

2026-07-01: Eight more state pages and six instruments: coverage reaches 45 of 51 jurisdictions

Added deep-dive pages for North Dakota, Ohio, Oklahoma, Rhode Island, Wisconsin, and Wyoming, plus two no-instrument state pages for South Carolina and Tennessee. Six instruments enter the dataset. North Dakota contributes the court system's Public Notice on the Use of Artificial Intelligence (June 1, 2026), practitioner-facing guidance that permits AI while leaving full responsibility for filings, fabricated citations included, with the attorney or self-represented litigant. Ohio contributes the Board of Professional Conduct's April 2026 guide, Artificial Intelligence for Lawyers and Judicial Officers, carried with its own framing intact: nonbinding staff guidance that expressly does not reflect the views of the Board or the Supreme Court of Ohio. Oklahoma contributes Rule 1.17 of the Rules of the Oklahoma Court of Criminal Appeals (2026 OK CR 7, effective February 18, 2026), which requires every AI-produced or AI-modified portion of an OCCA filing to be verified as accurate by a person responsible for the document, and which is the first criminal-appellate AI rule recorded in this tracker. Rhode Island contributes its Supreme Court's order of June 17, 2026, tracked as a court rule with dual character: it amends the comment to the professional-conduct competence rule, a binding conduct-rule comment amendment, and it adopts advisory Interim Guidelines for the Ethical Use of Generative AI for lawyers and judicial officers by court order. Wisconsin contributes the State Bar ethics counsel's official Ethical Dilemma column on generative AI (Wisconsin Lawyer, October 4, 2024), and Wyoming contributes Bar Counsel Mark W. Gifford's article in the February 2024 Wyoming Lawyer, each tracked as bar-staff guidance. The South Carolina and Tennessee pages document dated, verified negatives against the indexes we inspected, alongside each state's judiciary, bar, and case activity. State page coverage now stands at 45 of 51 jurisdictions, and the tracker lists 53 instruments.

2026-06-13: Updated: California COPRAC practical guidance (2026 revision)

Confirmed that California's Practical Guidance for the Use of Generative AI was revised in 2026 to add duties for agentic AI at the California Supreme Court's request, with State Bar Board of Trustees approval reported in May 2026. The separate proposed binding rule amendments remain in the pipeline and are still listed as proposed, not law.

2026-06-13: Added: Florida Supreme Court statewide AI filing rule (effective June 15, 2026)

The Florida Supreme Court amended Rule of General Practice and Judicial Administration 2.515(d)(2) (Case No. SC2026-0673, approved May 28, 2026, effective June 15, 2026): by signing a filing, the signer certifies that the legal authorities cited exist and are accurately cited, with express sanctions authority. A companion administrative order preempts circuit-level AI orders to create one statewide standard. Added as a court rule; the tracker now lists 34 instruments.

2026-06-13: Added: Illinois ARDC Guide to Implementing AI

Added the Attorney Registration and Disciplinary Commission's official Illinois Attorney's Guide to Implementing AI (October 24, 2025), which complements the Illinois Supreme Court AI policy with a classify-the-data, identify-the-tool, evaluate-the-safeguards framework plus a Practice Resource Kit. Tracked as guidance, distinct from a formal opinion.

2026-06-10: Watching: California proposed AI rule amendments

COPRAC's proposed amendments to Rules 1.1, 1.4, 1.6, 3.3, 5.1, and 5.3 finished public comment May 4, 2026. If adopted, these would be the first binding state disciplinary rules written specifically for AI. Listed as proposed, not law.

2026-06-10: Tracker launched with 32 verified instruments

Initial dataset: 16 formal ethics opinions (including ABA Formal Opinion 512), 10 guidance documents, 3 reports, 2 court rules/policies, and 1 pending rule proposal, every entry verified against its primary source. Honest labeling distinguishes formal opinions from guidance, reports, and court rules.

2026-06-10: New York 22 NYCRR Part 161 in effect

New York's system-wide court rule on AI use took effect June 1, 2026. It is permissive: no system-wide disclosure duty, with an optional Appendix A certification model rule for individual courts. Tracker entry and the court-disclosure explainer updated.

For licensed attorneys and firm operators. This site is legal information, not legal advice, and no attorney-client relationship is formed by using it. Rules change; verify against the primary sources linked on every page and consult a licensed attorney in your jurisdiction before acting.