The Legal AI State Bar Enforcement Gap Report 2026: How State Disciplinary Authorities Are — and Are Not — Investigating AI-Related Competence and Candor Complaints Filed in the First Half of 2026

Executive Summary

Across all 50 states and the District of Columbia, state bar authorities have generated an unprecedented volume of guidance addressing attorney use of artificial intelligence tools — yet formal disciplinary enforcement remains remarkably sparse, creating a regulatory gap with significant practical and liability implications for practitioners. Drawing on publicly available disciplinary filings, bar opinion activity, and court sanctions reporting through June 30, 2026, this briefing outlines the contours of that gap, maps what enforcement has actually occurred, and assesses what trajectory the second half of 2026 is likely to take.

A critical methodological note upfront: state bar disciplinary proceedings are, in most jurisdictions, confidential until a final finding of misconduct is issued and published. This means the true volume of open AI-related investigations is almost certainly substantially higher than publicly available data reflects. What follows is an analysis built from court dockets, published bar opinions, press accounts, and the narrow category of public disciplinary records — and should be read with that limitation explicitly in mind.

Methodology

Methodology note: This briefing synthesises public reporting, published rules and practitioner commentary. Figures marked as illustrative are estimates for discussion, not survey or test results. Sources include published disciplinary orders, ethics opinions and guidance, court sanction orders referencing AI-generated content, and published surveys such as the ABA's 2025 Legal Technology Survey and the 2026 Wolters Kluwer Future Ready Lawyer survey. Counts of states and orders are approximate, not the product of a systematic coding exercise.

AI-related conduct can be grouped into four operational types: competence failures (submission of AI-generated content without adequate review, including hallucinated citations); candor violations (affirmative misrepresentation to tribunals about the use or nature of AI-generated content); supervision failures (inadequate oversight of subordinate attorneys or nonlawyer staff using AI tools); and unauthorized practice adjacents (AI platform configurations that may have enabled nonlawyer actors to provide legal advice without supervision).

Jurisdictions can be grouped into four tiers: (A) formal ethics opinion issued; (B) informal guidance only — staff memos, FAQ documents, or CLE advisories without binding opinion status; (C) no published guidance but demonstrable enforcement action; and (D) no published guidance and no documented enforcement action.

The Guidance Landscape: A Taxonomy of Activity Without Accountability

As of June 2026, a growing number of states have issued formal ethics opinions specifically addressing AI — meaningful growth since end of 2024, but still leaving most jurisdictions operating on informal guidance or silence. California's State Bar issued Practical Guidance for the Use of Generative AI in November 2023, addressing competence, confidentiality, and supervision obligations under Rules 1.1, 1.6, and 5.3, and has since proposed AI-specific amendments to its Rules of Professional Conduct. Florida's Bar issued Ethics Opinion 24-1 in January 2024, addressing among other things client informed consent before confidential information is disclosed to third-party AI tools. New York, which produced a preliminary report in 2024, had not as of June 2026 converted that report into a binding formal opinion, instead operating through committee-level advisories — a distinction that has real enforceability consequences.

Texas, which adopted a technology competence mandate in its 2020 rules revision, addressed generative AI in Professional Ethics Committee Opinion 705 (February 2025), which applies existing competence standards. Illinois, Michigan, and Ohio remain in the informal guidance tier. A number of states — including Wyoming, North Dakota, and several smaller New England jurisdictions — had issued no documented guidance in any form as of the review period.

The pattern that emerges is troubling for consistency purposes: a practitioner in California using Harvey AI to draft a motion faces a clearer articulated standard than a practitioner in Wyoming using the same tool. Both face identical underlying professional responsibility obligations under Model Rules analogues, but only one operates in a jurisdiction where the bar has affirmatively staked out what compliance looks like.

Enforcement Findings: The Gap Is Wide

Against this guidance backdrop, documented formal disciplinary action attributable to AI use remains extraordinarily thin. Only a handful of published disciplinary orders through June 2026 explicitly reference AI-generated content as a triggering or contributing factor. This figure does not include court-imposed sanctions, which represent a parallel and in some respects more active enforcement mechanism.

The most consequential enforcement actions visible in public records have occurred at the court sanction level, not the bar level. The lineage runs from Mata v. Avianca (S.D.N.Y. 2023), where Judge Castel sanctioned attorneys from Levidow, Levidow & Oberman for submitting ChatGPT-hallucinated citations, through a series of 2024 and 2025 cases in which courts in Texas, Florida, and California imposed monetary sanctions and mandatory disclosure requirements. Several of those sanctioned attorneys were referred to state bars, but as of June 2026, few published disciplinary outcomes attributable to those court referrals had emerged — suggesting either that investigations remain open and confidential, that referrals did not result in charges, or some combination.

The most common triggering fact pattern across both court sanctions and the narrow set of bar actions is the submission of hallucinated case citations to a tribunal — a candor violation under Rule 3.3 — in which the attorney either affirmatively denied AI use when questioned or made no adequate inquiry into the citations' validity before submission. A secondary but growing pattern involves supervision failures: senior attorneys whose junior associates or paralegals used AI tools to generate work product that was filed without meaningful partner review. This pattern is likely to drive the next wave of formal disciplinary exposure as bars work through the supervisory liability framework under Rule 5.3.

In-house counsel exposure diverges meaningfully from outside counsel exposure in ways that current bar guidance largely fails to address. In-house attorneys typically do not file in tribunals and are therefore less exposed to the candor/hallucination fact pattern that has driven most visible enforcement. Their AI-related risk concentrates instead in the competence and confidentiality domains — specifically, the use of AI platforms without adequate data governance controls, potentially running afoul of Rule 1.6's confidentiality obligations and enterprise data security duties. We identified no state bar guidance targeting in-house counsel's AI obligations as distinct from outside counsel through June 2026, and no published disciplinary action against an in-house attorney for AI-related conduct during the review period. This does not mean zero investigation — again, confidentiality rules obscure the real picture.

Trajectory for H2 2026

Several indicators suggest that formal enforcement will increase materially in the second half of 2026, though likely not in proportion to the volume of guidance issued.

First, California's OCTC (Office of Chief Trial Counsel) can pursue AI-related complaints under existing competence and candor rules without waiting for new AI-specific rules. California's sheer bar membership — nearly 200,000 active attorneys (and more than 280,000 total licensees) — means that even a low complaint-to-action conversion rate generates meaningful enforcement numbers.

Second, mandatory AI disclosure rules adopted in several federal district courts in 2024 and 2025 — including a Northern District of Texas local rule (effective September 2024) and individual judges' standing orders in districts such as S.D.N.Y. and C.D. Cal. — have created a documentary record that makes referral and bar action more mechanically straightforward than in cases where AI use must be inferred.

Third, the proliferation of tools like Clio's AI assistant, Thomson Reuters' CoCounsel, and LexisNexis's Lexis+ AI into small and mid-size firm practice — populations historically associated with higher rates of disciplinary complaints — will expand the pool of potential triggering incidents.

The enforcement gap will likely narrow, but it will narrow unevenly: California, Florida, and New York will generate the bulk of visible action. Practitioners in guidance-silent jurisdictions will face a false sense of security — one that the second half of 2026 is likely to begin correcting through the uncomfortable mechanism of being first examples rather than informed adopters.

This briefing is based on publicly available data through June 30, 2026. Disciplinary proceedings that remain confidential under applicable state rules are not reflected. Readers should consult jurisdiction-specific counsel regarding current standards.