State Bar AI Ethics Guidance: What the Rules Actually Require

Generative AI has not created new ethical duties for lawyers. It has stress-tested the duties that already existed: competence, confidentiality, candor, supervision, and reasonable fees. The question for a managing partner is not whether the rules apply, but how a particular state has chosen to articulate them, and what to do if your state has said nothing at all.

This is a practical survey of the state bar AI ethics guidance issued so far, the points on which the states agree, and how a firm in a jurisdiction without guidance should orient itself. The short version: the differences between states are smaller than they appear, and a firm that follows the American Bar Association's framework will satisfy almost every state that has spoken.

The federal baseline: ABA Formal Opinion 512

In July 2024 the ABA Standing Committee on Ethics and Professional Responsibility issued Formal Opinion 512, "Generative Artificial Intelligence Tools." It was the ABA's first formal opinion on generative AI, and it remains the most complete single statement of the duties at issue.

Opinion 512 does not announce a new rule. It maps generative AI onto existing Model Rules: competence (Rule 1.1), confidentiality (Rule 1.6), communication (Rule 1.4), candor toward the tribunal (Rule 3.3), supervisory duties (Rules 5.1 and 5.3), and reasonable fees (Rule 1.5). Its operative themes are worth committing to memory, because they recur in nearly every state document that follows:

• Lawyers must develop a reasonable understanding of the capabilities and limitations of the tools they use, including the tendency of these systems to fabricate, or "hallucinate," authority.
• Inputting client information into a tool that may use it for training or expose it to third parties can implicate the duty of confidentiality, and may require client-informed consent.
• A lawyer cannot abdicate independent professional judgment to a tool. Output must be reviewed and verified.
• Billing must reflect work actually performed. A lawyer who bills hourly cannot bill for hours the tool saved.

Because Opinion 512 is comprehensive and rule-anchored, it functions as the de facto national standard. Several state documents predate it and read as narrower first drafts of the same ideas.

The states that have spoken

A meaningful and growing number of jurisdictions have issued formal opinions, court guidance, or practice notes. The following are among the most cited, and together they show the shape of the consensus.

California. The State Bar of California was early, issuing its Practical Guidance for the Use of Generative Artificial Intelligence in the Practice of Law in November 2023, before the ABA opinion. It is framed as principles and recommended practices rather than a numbered ethics opinion, and it leans heavily on confidentiality and on not charging clients for time the lawyer did not spend.

Florida. The Florida Bar issued Ethics Opinion 24-1 in early 2024. It addresses a lawyer's use of generative AI, with attention to protecting confidentiality, maintaining oversight of the work product, and disclosure where AI-related charges are passed to the client.

New Jersey. New Jersey's guidance came from the Judiciary rather than a bar ethics committee. In January 2024 the Supreme Court issued preliminary guidelines on the use of AI by attorneys, emphasizing that AI does not change a lawyer's existing obligations and that the duty to verify AI-generated information rests with the lawyer. Notably, the guidelines do not impose a general duty to disclose AI use in filings.

New York. The New York State Bar Association Task Force on Artificial Intelligence issued a substantial report and recommendations in April 2024, surveying the technology and the professional-responsibility issues it raises. New York committees have continued to issue more targeted guidance, including on using AI to record and transcribe client meetings, where consent and confidentiality are central.

Pennsylvania and Philadelphia. The Pennsylvania Bar Association and the Philadelphia Bar Association jointly issued Joint Formal Opinion 2024-200 in 2024. It is unusually thorough on rule-mapping, identifying obligations under competence (1.1), communication (1.4), confidentiality (1.6), conflicts (1.7), meritorious claims (3.1), candor (3.3), supervision (5.1), unauthorized practice (5.5), and misconduct (8.4).

Kentucky. The Kentucky Bar Association issued Ethics Opinion KBA E-457 in March 2024. It frames staying current on AI as part of the competence obligation, and takes a practical line on disclosure: routine use generally need not be disclosed to clients, but disclosure is appropriate when AI work is effectively outsourced or when AI costs are billed to the client.

District of Columbia. The D.C. Bar issued Ethics Opinion 388, "Attorneys' Use of Generative Artificial Intelligence in Client Matters," in 2024. It is direct about the technology, warning that these tools generate content statistically rather than retrieve it and therefore "hallucinate," and that lawyers must verify citations and analysis and protect client information from tools that retain inputs.

North Carolina. The North Carolina State Bar adopted 2024 Formal Ethics Opinion 1, "Use of Artificial Intelligence in a Law Practice." It tracks the now-familiar quartet: competence (1.1), confidentiality (1.6), supervision (5.1 and 5.3), and reasonable fees (1.5).

Texas. The State Bar of Texas Professional Ethics Committee issued Opinion 705 in early 2025, addressing generative AI in legal practice and stressing independent verification of AI output rather than blind reliance.

Virginia. Virginia has produced guidance on the fee question in particular, including Legal Ethics Opinion 1901 addressing reasonable fees for AI-assisted work. The thrust is that efficiency gains from AI bear on what a lawyer may charge, especially under hourly billing.

What the states agree on

Read side by side, these documents converge far more than they diverge. A firm that internalizes five points will be aligned with essentially all of them.

• Competence includes the tool. Lawyers must understand, at least in general terms, how generative AI works and where it fails. Several states tie this directly to the technology-competence duty.
• Confidentiality governs inputs. Putting client information into a consumer tool that trains on prompts or exposes them to third parties is the recurring danger. The duty is to vet the tool and, where appropriate, obtain informed consent.
• Verification is non-delegable. Every jurisdiction insists the lawyer, not the model, is responsible for the accuracy of what gets filed or sent. Hallucinated authority is the paradigm hazard.
• Supervision extends to AI. Rules 5.1 and 5.3 require partners and supervising lawyers to ensure that associates, staff, and the tools they use comply with the rules. A firm needs a policy, not just individual good intentions.
• Fees must be honest. A lawyer cannot bill hourly for time a tool eliminated. Charges for AI tools must be reasonable and, in many states, disclosed when passed to the client.

The genuine differences are at the margins: whether guidance comes from a bar committee or a court, how prescriptive the document is on disclosure, and how each state handles billing for non-hourly or flat-fee arrangements. None of these splits should change a sound firm-wide policy.

Why verification is the duty with teeth

The disciplinary risk is not theoretical, and it does not depend on your state having issued guidance. The cautionary case is Mata v. Avianca, Inc., 678 F. Supp. 3d 443 (S.D.N.Y. 2023). In June 2023, Judge P. Kevin Castel sanctioned plaintiff's counsel after they filed a brief containing citations to cases that did not exist, generated by ChatGPT and never verified. The court imposed a $5,000 sanction under Rule 11 and found the lawyers had acted in bad faith in standing by the fabricated authorities once questioned.

The Second Circuit reinforced the point in Park v. Kim, 91 F.4th 610 (2d Cir. 2024), where the court referred an attorney to its grievance panel after she cited a nonexistent case produced by ChatGPT. The lesson from both is the same, and it is the lesson every state opinion teaches: the tool's confidence is not evidence, and the signature on the filing is yours. Courts in numerous jurisdictions have since issued standing orders on AI use, which means the practical exposure runs through the courtroom regardless of what your bar has published.

If your state has not issued guidance

Most states have not yet published a dedicated generative-AI ethics opinion. That is not a gap in your obligations. The duties of competence, confidentiality, candor, supervision, and reasonable fees apply in every jurisdiction under its own rules of professional conduct, which are themselves modeled closely on the ABA Model Rules.

The practical answer is straightforward: treat ABA Formal Opinion 512 as the operative standard. A firm that follows 512 will satisfy the substance of every state opinion surveyed here, because those opinions are applications of the same Model Rules. Concretely, a small or mid-sized firm should:

• Adopt a written AI-use policy that names approved tools and prohibits entering client-identifying information into consumer tools that train on inputs.
• Require independent verification of every citation, quotation, and factual assertion that originates with or passes through an AI tool, with a human sign-off before filing or sending.
• Assign supervisory responsibility under Rules 5.1 and 5.3, so the obligation rests with a named person, not the firm in the abstract.
• Align billing practices with the duty of reasonable fees, and disclose AI-related charges that are passed through to clients.
• Check the local rules and standing orders of the courts you appear before, which increasingly speak to AI even where the bar has not.

Watch your own jurisdiction, because the roster is expanding quickly and new opinions tend to refine rather than reverse the consensus. Until then, the safest and most defensible posture is to read 512, write it into a policy, and verify everything.

This is general information for lawyers and law-firm leaders, not legal advice, and it does not create an attorney-client relationship. The authorities are cited so you can read them yourself.