Hold a short session on your written policy: approved tools, information that never goes in, how to verify citations, and who approves client-facing use. Include nonlawyer staff, use anonymized exercises, and repeat the session whenever you add a tool or your state issues new guidance. Write down who attended and when.
Start with the policy, then teach it
Training has nothing to teach until the firm has decided what is allowed. Write one page that lists the approved tools and what each may be used for, the information that never goes into any of them, who may approve a tool for client work and who answers questions. The session walks through that page with examples from your own practice areas.
The ethics rules put this on managers. ABA Model Rule 5.3 requires a partner or lawyer with comparable managerial authority to make reasonable efforts to ensure the firm has measures giving reasonable assurance that nonlawyers act compatibly with the lawyers' obligations1. A receptionist who pastes a voicemail transcript into a chatbot is a nonlawyer using a tool, which is why paralegals, assistants and front-desk staff attend.
Exercises that change behavior
Use matters stripped of names. One exercise gives each person an AI draft with invented citations hidden in it and asks them to find every one. The point lands because the cases are real: in Wadsworth v. Walmart, attorneys filed motions citing nine cases, and eight did not exist2. In Mata v. Avianca the court found the lawyers had abandoned their responsibilities when they submitted non-existent opinions and then stood by them after the court questioned them3.
A second exercise has people rewrite a prompt so it carries no client information and still gets a useful answer. A third has them read one vendor's terms and answer four questions: where inputs are stored, who can see them, whether they train models and how long they are kept.
What lawyers must understand
California's 2026 guidance describes two related duties of competence: a baseline understanding of the system's capabilities, data sources, limitations and risks, and independent judgment in reviewing, verifying and correcting outputs4. ABA Formal Opinion 512 says lawyers need not become experts but must have a reasonable understanding of a tool's capabilities and limitations5. Teach both halves: how the tool fails, and how a person catches the failure.
Keep it current
Repeat a short refresher when you approve a new tool, and send a one-paragraph note when your state bar or a court you appear in issues guidance. Keep the sign-in sheet and the policy version each person was taught. The ABA opinion also says a lawyer may not charge a client to learn how to use an AI tool the lawyer will regularly use5, so the firm absorbs the cost of its own training time.
We are not lawyers. Read the rule yourself or ask your counsel how it applies to you.
People also ask
Do I need to tell clients I use generative AI?
Sometimes. You must disclose if a client asks or requires it, and you need informed consent before client information goes into a self-learning tool. Internal use with no client information may need none. Check your state's opinion.
What is the best AI for lawyers?
No single product is best, and we have not tested enough to rank them. Pick by task and teach staff the approved list, and read a vendor's confidentiality terms before the tool sees any client file.
What kinds of AI tools do lawyers use?
General assistants, legal research platforms, document tools and front-office tools such as AI receptionists. Each has different data terms, so train staff on the kinds you have approved, not on AI in general.
Can AI handle intake for a law firm?
It can collect a name, a number and the matter type and pass them to your team, while saying it is an AI assistant. Staff should know it never advises callers or accepts a matter.
Is Claude or ChatGPT better for lawyers?
Neither is certified for legal work. Compare their terms on the plan you would buy, including training on inputs and retention, and have staff test both on an exercise that contains no confidential facts.
Sources
- American Bar Association. Rule 5.3: Responsibilities Regarding Nonlawyer Assistance (ABA Model Rules of Professional Conduct). American Bar Association, 2024. americanbar.org/groups/professional_responsibility/publications/model_
- U.S. District Court for the District of Wyoming. Wadsworth v. Walmart Inc., No. 2:23-CV-118-KHR, Order on Sanctions and Other Disciplinary Action (Doc. 181). U.S. District Court for the District of Wyoming (RECAP/CourtListener copy of the filed order), 2025-01-22. storage.courtlistener.com/recap/gov.uscourts.wyd.64014/gov.uscourts.wy
- Judge P. Kevin Castel, U.S. District Court, S.D.N.Y.. Mata v. Avianca, Inc., No. 22-cv-1461 (PKC), Opinion and Order on Sanctions (Doc. 54). U.S. District Court for the Southern District of New York (RECAP/CourtListener copy of the filed order), 2023-06-22. storage.courtlistener.com/recap/gov.uscourts.nysd.575368/gov.uscourts.
- State Bar of California, Committee on Professional Responsibility and Conduct. Practical Guidance for the Use of Generative Artificial Intelligence in the Practice of Law (Executive Summary). State Bar of California, 2026-05-14. calbar.ca.gov/Portals/0/documents/ethics/Generative-AI-Practical-Guida
- ABA Standing Committee on Ethics and Professional Responsibility. ABA Formal Opinion 512: Generative Artificial Intelligence Tools. American Bar Association, 2024-07-29. americanbar.org/content/dam/aba/administrative/professional_responsibi