ABA Formal Opinion 512: AI Ethics for Lawyers

Before anyone at your firm pastes a client’s facts into an AI tool or turns on an AI agent, know what ABA Formal Opinion 512 requires: your existing legal ethics duties apply, and the American Bar Association names seven.

Key Takeaways

• ABA Formal Opinion 512 (July 29, 2024) applies seven existing ethics duties to generative AI, from competence and confidentiality to candor and reasonable fees (ABA).
• Client consent is required before matter information goes into a self-learning AI tool, and “general, boiler-plate provisions” in an engagement letter are not enough (ABA).
• California’s 2026 guidance says lawyers must not let agentic AI give legal advice or file pleadings “without meaningful lawyer supervision and review” (State Bar of California).

On this page: Duties · Consent · States · AI agents · Checklist · Intake · FAQ

ABA Formal Opinion 512 applies seven existing duties to AI

Opinion 512, from the ABA’s ethics committee, covers generative AI (GAI): tools that create text, images, audio, video or code from a prompt. It applies existing ethical obligations under the Model Rules (ABA):

Its fee examples: 15 minutes spent entering information to draft a pleading can be billed, plus review time. A grammar checker is overhead; a per-use service reviewing thousands of contracts for one client can ordinarily be billed at actual cost. The same flat fee for much faster AI-assisted work “may be unreasonable under Rule 1.5” (ABA).

Client consent depends on what goes into the tool

Opinion 512 requires lawyers to get informed client consent before client information goes into a self-learning tool, one that keeps learning from what users enter. Brainstorming without client information needs none.

The client confidentiality risk runs inside the firm too: the tool could surface one client’s facts to a lawyer on another matter, even one behind an ethical wall. Consent must explain why the tool is used, its benefits and risks, and boilerplate in an engagement letter “is not sufficient.” Also read each tool’s “Terms of Use, privacy policy, and related contractual terms” (ABA).

Our take: use business plans that don’t train on your inputs, and keep client details out of prompts unless the client agreed. Our guide to AI tools for law firm marketing lists which ChatGPT and Claude plans train on inputs.

Bar guidance in four states mostly matches Opinion 512

Florida, California, New York and Texas largely line up with Opinion 512. Your state’s version governs; the Model Rules “serve as models for the ethics rules of most jurisdictions” (ABA).

A May 2025 New York City Bar Association roundup of a dozen jurisdictions flags one outlier: West Virginia, which it summarizes as concluding “that an attorney should receive client permission to use generative artificial intelligence before using it” (NYC Bar).

AI agents add access, autonomy and new ways to leak data

An AI agent carries out multistep tasks without a person prompting each one; California’s examples include “autonomously facilitating client intake” (State Bar of California). Opinion 512 doesn’t discuss agents, but its duties apply to the three risks they add:

1. Standing access. Agents can reach email, document systems and “client files,” so lawyers “must carefully evaluate and limit the scope of such access,” California says.
2. Actions without review. Don’t let agents send client information out on their own, “including automated communications, filings, or data transfers, without appropriate safeguards and human review.”
3. Hidden instructions. The OWASP Gen AI Security Project warns that “indirect prompt injections occur when an LLM accepts input from external sources, such as websites or files” (OWASP). A line planted in an email could steer an agent with access to your files.

Grant the least access the job needs: a tool that summarizes email “may only require the ability to read emails,” and a person should “approve high-impact actions before they are taken” (OWASP).

We build our own AI this way: our daily ad review has Claude propose changes, code checks each against the live account, and nothing changes until a person applies it. Our site’s chat sends transcripts to a person and reviews finished chats every 15 minutes (our setup, October 2026). The case management software we are building keeps AI off until the firm accepts the provider’s data-retention terms.

A checklist for the ethical use of AI in a law firm

Most steps come from Opinion 512. Our take: a small firm can do steps 1, 4 and 6 this week; vendor review takes longer.

1. List every AI tool, including AI in your CRM and phone system, and who uses each.
2. Vet AI service providers with 512’s questions: Is confidentiality enforceable? Will you hear of a breach or a legal demand for client data? Does the vendor keep your data or claim rights to it? What liability limits apply?
3. Keep inputs out of model training. California: “Reasonable efforts require more than reliance on generalized marketing assurances.”
4. Get informed consent before client information enters a self-learning tool.
5. Limit agents to the access they need, with a person approving anything sent outside the firm.
6. Review every output before it reaches a client, court or the public; check court rules on AI disclosure.
7. Explain any AI charge before billing it, preferably in writing.
8. Train everyone and mark AI-made material in the file “so future users understand potential fallibility of the work.”
9. Recheck tools after updates or before new tasks, as California asks.

Marketing and intake tools follow the same rules

Someone who types a DUI arrest or custody dispute into your website chat may be a prospective client, whose information Rule 1.18(b) protects, as 512 notes.

Florida’s Opinion 24-1 addresses intake chatbots directly (The Florida Bar):

• A lawyer “must inform prospective clients that they are communicating with an AI program and not with a lawyer or law firm employee.”
• An “overly welcoming” chatbot that gives legal advice risks a client relationship “without the lawyer’s knowledge.”
• Screening questions can limit the bot if a person already has a lawyer.

The NYC Bar’s opinion says a lawyer “must adequately supervise the chatbot” and must not use AI “to pose as a real person to communicate with prospective clients” (NYC Bar). Our take: put the AI disclosure in the chat’s first message, not a footer.

Our AI receptionist guide covers chatbot setup. For a second look at your intake, start with a Law Firm Growth Evaluation.

Frequently asked questions about ABA Formal Opinion 512

Is ABA Formal Opinion 512 binding?

Not by itself. It interprets the ABA Model Rules; your state’s rules govern. Florida’s 24-1 even opens with “Advisory ethics opinions are not binding” (The Florida Bar).

Do lawyers have to tell clients they use AI?

Sometimes. Opinion 512 requires it when, for example, a client asks, client information goes into a self-learning tool, AI affects the fee, or its output shapes a significant decision. Otherwise, “client disclosure may be unnecessary” (ABA).

Treat AI like an assistant a lawyer supervises

Florida and Texas both liken relying on AI to relying on “inexperienced or overconfident” nonlawyer assistants, and both hold the lawyer responsible for the work. More on AI: our guide to AI for law firms.

This post is general information, not legal advice.

Sources

1. American Bar Association, “Formal Opinion 512: Generative Artificial Intelligence Tools” (July 29, 2024), https://www.americanbar.org/content/dam/aba/administrative/professional_responsibility/ethics-opinions/aba-formal-opinion-512.pdf, checked October 21, 2026.
2. The State Bar of California, “Practical Guidance for the Use of Generative Artificial Intelligence in the Practice of Law” (2026), https://www.calbar.ca.gov/sites/default/files/portals/0/documents/ethics/Generative-AI-Practical-Guidance.pdf, checked October 21, 2026.
3. American Bar Association, “Model Rules of Professional Conduct,” https://www.americanbar.org/groups/professional_responsibility/publications/model_rules_of_professional_conduct/, checked October 21, 2026.
4. The Florida Bar, “Ethics Opinion 24-1” (January 19, 2024), https://www.floridabar.org/etopinions/opinion-24-1/, checked October 21, 2026.
5. New York City Bar Association, “Formal Opinion 2024-5: Generative AI in the Practice of Law” (August 2024), https://www.nycbar.org/reports/formal-opinion-2024-5-generative-ai-in-the-practice-of-law/, checked October 21, 2026.
6. Texas Center for Legal Ethics, “Opinion 705” (Professional Ethics Committee for the State Bar of Texas, February 2025), https://www.legalethicstexas.com/resources/opinions/opinion-705/, checked October 21, 2026.
7. New York City Bar Association, “Current Ethics Opinions and Reports Related to Generative Artificial Intelligence” (May 2025), https://www.nycbar.org/wp-content/uploads/2025/05/20221469-AnalysisEthicsGuidanceAI.pdf, checked October 21, 2026.
8. OWASP Gen AI Security Project, “LLM01:2025 Prompt Injection,” https://genai.owasp.org/llmrisk/llm01-prompt-injection/, checked October 21, 2026.
9. OWASP Gen AI Security Project, “LLM06:2025 Excessive Agency,” https://genai.owasp.org/llmrisk/llm062025-excessive-agency/, checked October 21, 2026.

Header image: AI-generated image.