State Bar Guidance on AI Use in Legal Practice: What to Know

Written by the Citation Safe Research Desk · Reviewed by Andy Gaber, Founder — September 3, 2026

Several state bars and the American Bar Association have issued formal guidance on generative AI use in legal practice. None of them prohibit AI-assisted drafting outright; all of them emphasize competence, confidentiality, and independent verification as the core duties that don't change just because a tool is new.

ABA Formal Opinion 512

The American Bar Association's Formal Opinion 512 (2024) addresses generative AI tools directly, reaffirming that the duty of competence (Model Rule 1.1) requires lawyers to understand the capabilities and limitations of any AI tool they use, and that the duty of candor to the tribunal is not satisfied by relying on unverified AI output.

California

The California State Bar issued practical guidance on generative AI use in late 2023, addressing confidentiality (caution against inputting client information into consumer-facing tools without safeguards), competence, and the need to independently verify AI-generated legal research before relying on it in any filing or client communication.

Florida

The Florida Bar's ethics opinion on generative AI addresses similar themes: confidentiality obligations when using third-party AI tools, supervision duties over AI-assisted work product, and billing transparency around AI-assisted efficiency gains.

New York

The New York State Bar Association's task force report on AI addressed generative AI's implications across competence, confidentiality, and the specific risk of hallucinated case law reaching a filing, citing the emerging public record of sanctions cases as a central concern for the profession.

The common thread across jurisdictions

Every state guidance document reviewed converges on the same practical requirement: independent verification of AI-generated legal research before it is relied upon in any filing, client advice, or professional representation. None treat "the AI tool told me it was accurate" as sufficient.

Expert perspective

Legal ethics scholars have noted that these opinions largely apply existing rules (competence, confidentiality, candor, supervision) to a new tool rather than creating novel AI-specific obligations — meaning attorneys who already take those duties seriously have most of what they need, provided they add an explicit verification step for AI-assisted research.

A common question

Does my state need to have issued specific AI guidance for these duties to apply?

No. Even without state-specific guidance, existing Model Rules on competence, confidentiality, and candor apply directly to AI-assisted work. State-specific opinions clarify how those existing duties apply; they do not create the underlying obligation.

How state guidance documents actually get used in practice

Most state bar AI guidance documents are advisory rather than binding disciplinary rules in themselves, but they matter in practice because bar disciplinary counsel and courts routinely cite them when evaluating whether an attorney's conduct met the applicable standard of care. An attorney who ignored published guidance directly on point is in a materially worse position when a hallucination incident is reviewed than one who can show some good-faith effort to follow it, even imperfectly.

Confidentiality: the guidance most attorneys underweight

Nearly every state guidance document spends as much space on confidentiality as on accuracy, and it is easy to focus only on the hallucination risk while overlooking it. Pasting client facts, even anonymized-seeming details, into a consumer-facing AI tool without checking the tool's data-retention and training-use policies can itself violate confidentiality duties independent of whether any citation in the output is accurate. Firms that have adopted enterprise-tier AI tools with contractual no-training guarantees are generally in a stronger position on this specific point than those relying on free consumer tiers.

Keeping up as guidance evolves

State guidance in this area has been revised more than once in some jurisdictions as AI tools and case law have both evolved quickly. Treat any single guidance document as a snapshot rather than a permanent rule, and check for updates at least annually, particularly if your state bar has an active technology or AI task force.

How to build a compliance file that actually helps you

Rather than treating state guidance as background reading, build a short internal memo mapping each guidance document's specific requirements (competence, confidentiality, supervision, candor) to a concrete step in your firm's own workflow. For a solo practitioner, this can be a single page. For a larger firm, it typically becomes part of the same manual that already covers conflicts checks and trust accounting. The goal is a document you could hand to a bar investigator or malpractice carrier that shows, concretely, how your practice implements the guidance rather than merely being aware of it.

What happens when guidance is ignored entirely

Bar disciplinary proceedings arising from AI-related conduct have, in the documented cases, treated a complete absence of any verification process more harshly than a good-faith process that still missed an error. The distinction matters: guidance documents are not primarily about achieving zero errors, they are about establishing a demonstrable standard of care that a reasonable attorney would be expected to follow given what is now widely known about AI hallucination risk.

Final takeaway

Read your own state's guidance at least once, even briefly, rather than assuming it says what you'd expect. Several attorneys in documented sanctions cases were surprised to learn after the fact that their state bar had already published guidance directly addressing the exact failure that got them sanctioned. Ignorance of published guidance has not been treated as a mitigating factor in any documented case reviewed for this piece.

For multi-jurisdictional practices, keep a simple internal reference table listing which states have issued specific generative-AI guidance and a one-line summary of any distinctive requirement, updated at least twice a year. This is a low-cost way to avoid being caught unaware by a state-specific rule that differs from your home jurisdiction's approach.

When in doubt about your own jurisdiction's specific position, a short call to your state bar's ethics hotline, where one exists, is a free and direct way to get a current answer rather than relying on a guidance document that may already be outdated.