State Bar AI Guidance: A 50-State Comparison

Over 30 states have issued formal guidance. The themes converge, but the details diverge — and the details are what get lawyers sanctioned.

---

The American approach to AI regulation in legal practice is, characteristically, fragmented. There is no single national standard. Instead, there is ABA Formal Opinion 512 providing a baseline, over 30 states issuing their own guidance, and individual courts layering additional requirements on top.

For lawyers practising across state lines — which in 2026 means most lawyers — this creates a compliance puzzle that requires careful navigation. Here is how I read the current landscape.

The National Baseline: ABA Opinion 512

Opinion 512, issued July 2024, applies existing Model Rules to AI use. It does not create new obligations. It interprets existing ones in the context of generative AI. The seven pillars are competence (Rule 1.1), confidentiality (Rule 1.6), communication (Rule 1.4), candor to the tribunal (Rules 3.1 and 3.3), supervision (Rules 5.1 and 5.3), and reasonable fees (Rule 1.5).

Every state bar guidance document I have reviewed is consistent with these pillars. The variation comes in emphasis, specificity, and — critically — whether the guidance is mandatory or advisory.

States That Have Gone Mandatory

Three states stand out for imposing requirements that go beyond guidance into mandate.

Pennsylvania (August 2024) mandates explicit disclosure of AI use in all court submissions. This is not a recommendation — it is a filing requirement. Pennsylvania also issued Joint Formal Opinion 2024-200 through the Pennsylvania and Philadelphia Bar Associations, providing detailed ethical guidance on AI use. If you file in Pennsylvania, disclosure is non-optional.

New York (Q3 2025) requires at least two annual Continuing Legal Education credits in practical AI competency. New York has taken the position that AI competence is not something lawyers can develop on their own — structured education is required, and it must be documented.

California (January 2025) requires multi-jurisdictional compliance for AI cloud tools. If your AI vendor processes data outside California, California rules still apply. This has significant implications for firms using cloud-based AI tools with servers in multiple jurisdictions.

These three states represent the leading edge. The rest of the country is moving in the same direction, but has not arrived yet.

States With Formal Opinions

A larger group of states has issued formal ethics opinions addressing AI — not mandatory requirements, but authoritative interpretations that disciplinary bodies will reference.

Florida (Opinion 24-1, January 2024) covers confidentiality, oversight, fee reasonableness, and advertising compliance. Florida's emphasis on advertising is notable — if you use AI to generate marketing content, Florida's opinion has specific implications.

Texas (Opinion No. 705, February 2025) provides specific guidance on verification requirements and confidentiality safeguards. Texas practitioners should read this alongside TRAIGA for a complete picture of their AI obligations.

Oregon (Opinion No. 2025-205) is particularly useful on billing ethics, addressing how AI-generated time savings should be reflected in client bills and when AI costs can be passed through to clients.

North Carolina (Formal Ethics Opinion 1, 2024) emphasises competence and supervision obligations — specifically, that supervising lawyers must understand AI well enough to oversee its use by others.

New York City Bar (Opinion 2025-6) addresses a specific emerging issue: using AI to record, transcribe, and summarise client conversations. This is ahead of most other guidance in addressing real-time AI applications in client interactions.

States Still Developing

Several states have established committees or working groups but have not yet produced formal guidance. Georgia's Special Committee on AI and Technology is examining whether current ethical rules are sufficient. Hawaii's Committee on AI and the Courts was established by the Chief Justice. Nevada has formed an advisory group.

A handful of states — North Dakota, Ohio, Oklahoma, Rhode Island, South Carolina, South Dakota — have not issued formal guidance at all as of this writing.

The absence of formal guidance does not mean the absence of obligations. ABA Model Rules, as adopted in each state, apply to AI use whether or not the state bar has issued specific guidance.

The Five Universal Themes

Despite jurisdictional variation, five themes appear in virtually every state's guidance.

Competence. Lawyers must understand AI tools well enough to use them competently. This is not about becoming a technologist — it is about understanding capabilities, limitations, and failure modes sufficiently to make informed decisions about when and how to use AI.

Confidentiality. Client information entered into AI systems must be protected. This requires understanding vendor data handling practices, processing locations, and terms of service. Using a consumer AI tool for client work without understanding these factors is a confidentiality violation in every jurisdiction.

Supervision. AI tools and their vendors are service providers under Model Rule 5.3, requiring appropriate oversight. Supervising lawyers must ensure that subordinate lawyers and staff comply with AI policies.

Verification. No state permits filing AI output without human review and verification. Every citation must be confirmed. Every factual claim must be checked. Every legal proposition must be validated.

Fee reasonableness. AI efficiency gains must be reflected honestly in billing. Charging for time the task "would have taken" without AI is impermissible. Learning time for general AI tools cannot be billed to clients.

Multi-Jurisdictional Strategy

For lawyers practising across states, the compliance strategy is straightforward in principle, complex in execution: comply with the most restrictive jurisdiction's requirements everywhere.

If any jurisdiction you practise in requires AI disclosure, disclose everywhere. If any jurisdiction requires AI-focused CLE, complete it. Maintain confidentiality practices that meet the highest applicable standard. Document compliance for every jurisdiction.

This approach costs more in the short term but eliminates the risk of jurisdiction-specific compliance failures. And it positions you well as other states inevitably move toward mandatory requirements.

The Trajectory

The pattern is clear. Guidance becomes formal opinion. Formal opinion becomes mandatory requirement. What is voluntary today will be compulsory within twelve to twenty-four months.

Pennsylvania's disclosure mandate, New York's CLE requirement, and California's multi-jurisdictional compliance rule are not outliers — they are previews. Every state bar is moving in this direction. The variable is timing, not destination.

Lawyers who build compliance infrastructure now will be positioned for tightening requirements. Those who wait will face retrofitting costs, compliance gaps, and — in the worst case — disciplinary exposure for practices that were technically permissible when adopted but are sanctioned by the time they cause problems.

Build the infrastructure now. The regulatory environment is moving in one direction only.