What does ABA Formal Opinion 512 say about AI tools?

ABA Formal Opinion 512 applies the ABA's existing ethics rules to generative AI. It asks three things of a lawyer. Understand the tool well enough to know what it can and cannot do. Make reasonable efforts to protect the client's information. And get the client's informed consent before entering client information into a self-learning tool that could disclose it. The consent rule changes what a lawyer must do, because it applies when a tool learns from what users type or say.

The opinion starts from the duty of reasonable efforts

The opinion's confidentiality section rests on Model Rule 1.6(c), which requires a lawyer to make "reasonable efforts to prevent the inadvertent or unauthorized disclosure of, or unauthorized access to, information relating to the representation of the client" ¹. The ABA's earlier opinion on electronic communications says that what counts as reasonable is "not susceptible to a hard and fast rule, but rather is contingent upon a set of factors" ². So the opinion does not list one approved tool. It asks you to judge the risk of each tool and its terms.

Lawyers must understand the tool

The competence part of the opinion is about understanding. Lawyers "must have a reasonable understanding of the capabilities and limitations of the specific GAI technology that the lawyer might use" ³. In practice that means knowing whether a tool keeps what you enter, whether it learns from it and how reliable its output is. The vendor's terms answer the first two questions.

Self-learning tools raise a confidentiality risk

The confidentiality concern is specific. The opinion says that self-learning generative AI tools "by their very nature, raise the risk that information relating to one client's representation may be disclosed improperly" ⁴.

That is why consent is central. The opinion concludes that informed consent is required before client information is entered into self-learning generative AI tools whose outputs could disclose it ⁵. The consent must be informed, so the lawyer explains the material risks before asking for it.

A general engagement-letter clause does not count as consent

A standard clause in an engagement letter is not enough. The opinion says that "merely adding general, boiler-plate provisions to engagement letters purporting to authorize the lawyer to use GAI is not sufficient" ⁶. The firm needs a specific conversation with the client and a record of it.

State rules differ, so the ABA opinion is not the last word

The ABA opinion is guidance. Your state's rules and its bar's opinions decide what binds you, and they vary in how firmly they put consent. Texas says a lawyer who puts confidential information into a generative AI tool should consider telling clients about the risks, and may need their consent ⁷. That is a softer duty than the ABA's requirement for self-learning tools. Read your own state's opinion before you rely on either.

It depends: what the tool does with your input

• A tool that learns from inputs. The ABA's consent rule applies. Get specific informed consent and record it.
• A tool that processes input and keeps nothing. The self-learning risk is smaller. You still need reasonable efforts and a written check of the terms.
• A tool whose terms you cannot read or get. You cannot show reasonable efforts without that information, so do not put client information in it.

What to do under Opinion 512

1. Ask the vendor whether the tool learns from what users enter. Get the answer in writing, and check the retention period and any human review.
2. If it learns, get specific informed consent. Explain the risk that later users may access the information, and keep a record.
3. If it does not, document the check. Save the terms, the settings you chose and the date you reviewed them.
4. Do not rely on the engagement letter alone. The opinion says a boilerplate clause is not enough.

For how these duties apply to dictation tools, see our guide on dictation for lawyers, privilege and confidentiality. The related questions cover whether cloud dictation waives attorney-client privilege, whether AI chatbot conversations are privileged and whether lawyers can use AI notetakers on client calls. For the broader rules, see attorney-client privilege and our overview of voice privacy and security.

This is a summary of an ethics opinion, not legal advice. Check the ABA's text and your state's rules before you rely on it for a matter.