AI ethics guide
ABA Formal Opinion 512, in plain English
Can a lawyer put client documents into an AI tool? What the American Bar Association's first formal opinion on generative AI says, and what to check before you do.
Not legal advice. This is a plain-English summary of published ethics guidance, written by the team that makes ClientSafeAI, and it hasn't been reviewed by a lawyer. Ethics opinions are advisory, and your own state's rules govern. Read the opinion itself, and ask your state bar's ethics hotline if you're unsure.
The short answer
Yes, but not casually. Formal Opinion 512, issued on July 29, 2024 by the ABA's Standing Committee on Ethics and Professional Responsibility, says the existing Model Rules already cover generative AI, mainly the rules on competence, confidentiality, communication and fees. Its sharpest line is about self-learning tools, the ones that learn from what users put in. Because their output could lead directly or indirectly to disclosing client information, the opinion says a client's informed consent is required before information relating to the representation goes into one.
A boilerplate clause in an engagement letter doesn't count as that consent. And if the tool doesn't learn from your input and doesn't expose it to anyone else, the analysis is different. That's why the first question about any AI tool is what happens to what you give it.
What the opinion covers, rule by rule
| Model Rule | What Opinion 512 says |
|---|---|
| 1.1 Competence | Understand the benefits and risks of the technology you use to deliver legal services. That includes a reasonable understanding of what a given AI tool can and can't do. You don't need to be an expert, but it's “not a static undertaking”: keep up as the tools change, and expect tools that become standard in practice to raise what competence means. |
| 1.6 Confidentiality | Informed consent before putting information relating to the representation into a self-learning tool. Self-learning tools raise the risk of improper disclosure “even if the tool is used exclusively by lawyers at the same firm.” Consent isn't needed if no information relating to the representation goes in, for example when using a tool only for idea generation. |
| 1.4 Communication | Even where Rule 1.6 doesn't require consent, consider whether the client should be told. Rule 1.4(a)(2) requires you to reasonably consult with the client about how their objectives will be met. |
| 1.5 Fees | If you bill hourly, bill actual time. You can charge for the time spent prompting a tool and reviewing its draft, but in most circumstances not for learning how to use it. Expenses must be reasonable and generally passed on at actual cost, and whether a tool is overhead or a billable cost depends on the tool and how it's used. |
| 3.1, 3.3 Candor | Nonexistent citations and inaccurate analysis are the lawyer's responsibility. Some courts now require lawyers to disclose AI use. Before anything goes to a court, review the output, citations included, and correct errors. |
| 5.1, 5.3 Supervision | Managerial lawyers must set clear firm policies on permissible AI use, and supervisors must make reasonable efforts to ensure lawyers and nonlawyers follow them. Vetting an AI vendor follows the outsourcing opinions: reference checks, credentials, how the vendor protects confidentiality, and its conflicts screening. |
What informed consent has to include
The opinion is specific. For consent to be informed, the client needs the lawyer's best judgment about why the AI tool is being used, specific information about the risk (including which kinds of client information will be disclosed and how others might use it against the client's interests), and a clear explanation of how the tool benefits the representation. Part of that is explaining the risk that later users of the tool could get access to the client's information.
Boilerplate waivers will not suffice as informed consent.
The Bar Examiner's summary of Formal Opinion 512
Before you put client information into an AI tool
A working checklist drawn from Opinion 512 and the state opinions below. It's a starting point for a firm policy, not a substitute for one.
- Read the terms of use and privacy policy. Find out who can access what you enter, whether the provider keeps it, and whether it trains on it. If you can't tell, ask a colleague or outside expert who can.
- Decide whether the tool is self-learning. If it learns from input or could surface it to other users, get the client's informed consent first, or keep client information out of it.
- Strip what you don't need. If no information relating to the representation goes in, the confidentiality analysis changes. Several state opinions say the same about anonymized input.
- Check where it runs and who else is on it. A tool shared with other firms or other clients carries different risks from one that runs only for you.
- Verify every citation and quotation before it goes into anything you file or send.
- Bill actual time. Settle up front how any AI costs will be handled, and don't bill for learning the tool.
- Write it down. A short firm policy on which tools are approved and for what, plus training for everyone who uses them.
- Check your state and your court. State opinions differ on when you must tell the client, and some courts have their own AI orders.
What your state says
Several state bars and courts have published their own guidance. They mostly agree with Opinion 512, but they differ on when consent or disclosure is needed. Each page below is written from that jurisdiction's own opinion and links to it.
Don't see your state? We only publish a page where there's a formal opinion or official guidance to summarize. Opinion 512 and your state's confidentiality rule are the place to start, and your state bar's ethics hotline can tell you whether anything newer applies.
An AI tool that answers the first question simply
A Private AI Workspace runs on a GPU server ordered for one firm. Your documents, the AI model and the database all stay on it, it doesn’t train on your files, and no other customer shares it. Answers cite the page they came from, so checking them is quick.
Sources
- ABA Standing Committee on Ethics and Professional Responsibility — Formal Opinion 512, Generative Artificial Intelligence Tools (PDF)
- ABA News — ABA issues first ethics guidance on a lawyer’s use of AI tools (July 29, 2024)
- The Bar Examiner (NCBE) — Generative Artificial Intelligence Tools, a summary of Formal Opinion 512
Spotted something out of date? Tell us and we'll fix it.
Not legal advice. This is a plain-English summary of published ethics guidance, written by the team that makes ClientSafeAI, and it hasn't been reviewed by a lawyer. Ethics opinions are advisory, and your own state's rules govern. Read the opinion itself, and ask your state bar's ethics hotline if you're unsure.