ABA 512 Compliance — What State Bars Will Ask You in 2026

In August 2024, the American Bar Association House of Delegates adopted ABA Resolution 512. The accompanying Formal Opinion 512 had been released by the ABA Standing Committee on Ethics and Professional Responsibility on July 29, 2024 — roughly one week earlier. Together they are the first national-level ethics articulation for generative AI in legal practice. In the same window, at least ten state and local bars have layered their own opinions on top. Most solo and small-firm shops we have audited still do not have a written firm AI policy at all.

The gap is the whole reason /legal/aba-512-audit exists. It walks a firm through what ABA 512 and the major state opinions actually require, scores the firm against a 12-point checklist, and outputs a policy draft the firm can take to its outside ethics counsel before signing. None of what follows is legal advice. Run the actual policy past a Lawyers' Professional Responsibility expert in your jurisdiction before you adopt it.

What ABA Resolution 512 and Formal Opinion 512 actually say

The ABA's posture in Formal Opinion 512 is that the existing Model Rules already cover generative AI use — there are no new rules, only existing rules applied to a new tool. The opinion walks through seven duties: competence (Rule 1.1), confidentiality (Rule 1.6), communication (Rule 1.4), supervisory obligations (Rules 5.1 and 5.3), responsibilities to the tribunal (Rule 3.3), reasonable fees (Rule 1.5), and the duty to comply with applicable law.

The two clauses to commit to memory are the verification clause and the supervision clause.

On verification, Comment 8 to Model Rule 1.1 — the technology-competence comment now adopted in 40+ states — requires lawyers to "keep abreast of changes in the law and its practice, including the benefits and risks associated with relevant technology." Formal Opinion 512 reads that against generative AI's well-documented hallucination problem and concludes that lawyers must independently verify any factual or legal output from a GAI tool before relying on it. The Mata v. Avianca sanction, the Park v. Kim sanction, and the half-dozen sanctions that have followed are the disciplinary record on what happens when lawyers do not.

On supervision, Rules 5.1 and 5.3 carry over directly. A managing partner is responsible for ensuring associates and non-lawyer staff (including AI tools used as if they were non-lawyer staff) operate within the rules. That obligation is non-delegable. The opinion is explicit that a firm cannot outsource its AI-supervision duty to the vendor.

The third sticky clause is on confidentiality. Model Rule 1.6 prohibits disclosing information relating to the representation without informed consent. Formal Opinion 512 treats the act of feeding client information into a self-learning GAI tool as a potential disclosure.

The state bars that have layered on top

Ten of the 50 state bars (plus DC and several judicial-conduct counterparts) have published formal opinions or written guidance since Q4 2023. The pattern across them is consistent — most cite the ABA opinion approvingly, then add jurisdiction-specific overlays.

California. The State Bar's Standing Committee on Professional Responsibility and Conduct issued Practical Guidance for the Use of Generative AI in the Practice of Law in November 2023 — the first major state-bar document on the topic. California's overlay is heavy on confidentiality and on billing. Time saved by AI cannot be billed as if it were lawyer time.

Florida. Florida Bar Ethics Opinion 24-1 (January 2024) is the most prescriptive state opinion to date. It mandates confidentiality protections, lawyer oversight, fee fairness, and advertising restrictions, and explicitly addresses chatbot-style client intake. Florida is also explicit on client notification: when the AI use is "extensive," the lawyer should consider informing the client.

New York. The New York State Bar Association published its Task Force on AI's Report and Recommendations in April 2024 — the most comprehensive jurisdiction-level treatment, 91 pages. NYSBA flagged vendor due diligence as the area where most firms fall short.

New Jersey. The NJ Supreme Court's Preliminary Guidelines (January 2024) is notable for being issued by the court rather than the bar — meaning judges are the audience for enforcement, and the guidelines have a near-rule-like quality. The NJ guidelines lean hardest on the verification duty and on candor to the tribunal.

Pennsylvania + Philadelphia. Joint Formal Opinion 2024-200 (May 2024) is heavy on the unauthorized-practice question and on the scope of client consent. PA treats the AI-as-non-lawyer-assistant framing under Rule 5.3 more strictly than the ABA.

Texas. Texas Opinion 705 (February 2025) is the newest of the major opinions. Texas's contribution is on the obligations of lawyers who use GAI tools provided by their employer — meaning associates and contract attorneys at firms where the tooling decision was made above their pay grade still bear individual verification responsibility.

What state bars are actually asking firms during oversight conversations

Disciplinary counsel and bar inquiry boards have started including AI questions in routine investigations and random audits. The questions are remarkably consistent:

• Does the firm have a written AI use policy?
• Who approved it and when?
• Which AI tools are covered? Which are explicitly prohibited?
• What is the protocol for verifying AI output before it is included in work product?
• How are AI prompts containing client information handled?
• What training have firm personnel received on the policy?
• What is the firm's incident response process if a confidentiality breach involving an AI tool is discovered?
• How does the firm describe its AI use to clients in engagement letters?
• How does the firm bill for AI-augmented work?
• What due diligence was done on the AI vendor before adoption?

A firm that cannot answer at least eight of these in writing is exposed. The audit at /legal/aba-512-audit is built to score the firm against the full list and surface the gaps.

The 12-point firm AI policy checklist

Every firm AI policy worth its name covers these twelve elements.

• Covered tools. Name the AI tools the firm uses, the tools it explicitly prohibits, and the procedure for adding a new tool to the approved list.
• Confidentiality rules. Define what categories of client information may be entered into which tools.
• Verification protocols. State the firm's standard for verifying AI output before it goes into work product or court filings.
• Supervision lanes. Define who in the firm is responsible for supervising AI use by associates, contract attorneys, paralegals, administrative staff, and outsourced support.
• Billing and disclosure. State how time saved by AI is treated in fee arrangements.
• Competence training. Require that any firm personnel using an AI tool receive documented training on the policy.
• Client notice. Define the threshold at which a client must be affirmatively notified of AI use.
• Prohibited uses. Name the categories of work the firm will not delegate to AI.
• Vendor due diligence. Document the diligence performed on each approved vendor.
• Data retention. Define how long AI prompts and outputs are retained, where they are stored, and how they are destroyed at end of representation.
• Incident response. Define the procedure when a suspected confidentiality breach involving an AI tool is discovered.
• Annual review. Schedule an annual review of the policy with documented sign-off by the managing partner.

What it costs to do this two ways

Drafting and reviewing a firm AI policy through outside counsel runs $2,000 to $5,000 for a sub-10-attorney firm. The math: 5-12 hours of partner time at $400-600/hr to draft, plus 2-4 hours of associate or paralegal time on the supporting checklists, plus 1-2 hours of outside-ethics-counsel review.

The free path is the audit at /legal/aba-512-audit. Run it against your existing posture, get the gap report, take the output template to your ethics counsel for the 1-2 hour review pass. Total cost: $0-$200 in counsel time on the review side.

What to do this week

If the firm has no written AI policy, run the audit today and get a draft on the managing partner's desk by Friday. If the firm has a policy that has not been reviewed since the Florida 24-1 release in January 2024, run the audit against the current policy — there is a meaningful chance it is missing the Texas Opinion 705 employer-tool clause or the Pennsylvania 2024-200 unauthorized-practice clarification.

The bar is not asking for perfection. It is asking for a written policy, evidence the policy was adopted, evidence personnel were trained on it, and evidence the firm verifies AI output before relying on it. The audit is built to produce all four.

Do the work this month. The disciplinary cases that will define this area through 2027 are being filed right now, and they are landing on firms that did not.