ABA Formal Opinion 512 Explained for Firms and Legal Departments

ABA Formal Opinion 512, issued July 29, 2024, is the ABA’s first formal ethics opinion on generative AI and, more than two years later, still the document most firms and legal departments start from when they write an AI policy. It creates no new rules; the 15-page opinion explains how existing Model Rules on competence, confidentiality, communication, meritorious claims and candor, supervision and fees apply. Here is what it says, and where its limits are.

What Is ABA Formal Opinion 512?

The opinion comes from the ABA Standing Committee on Ethics and Professional Responsibility. ABA opinions interpret the Model Rules, which states adopt in their own versions, so the opinion does not bind any lawyer directly. The committee’s own page notes that ABA formal opinions have been cited as persuasive. The opinion itself anticipates that the committee and state and local bar ethics committees will offer updated guidance as specific tools develop. The committee has issued a dozen opinions since, numbered 513 through 524, and none of them returns to generative AI, so 512 remains the ABA’s main word on the subject.

What Does Opinion 512 Say About Competence and Confidentiality?

On competence under Rule 1.1, the opinion says lawyers do not need to be technology experts but do need a reasonable understanding of the capabilities and limits of the tools they use. It is direct about why: it cites a Stanford study finding that leading legal research tools “hallucinate between 17% and 33% of the time.” The amount of checking it describes depends on the tool and the task. Its example is useful in practice: a lawyer who has tested a tool’s contract summaries on a smaller subset of documents may not need to review every summary by hand.

Confidentiality under Rule 1.6 is where the opinion is most specific. For tools that learn from what users enter, which the opinion calls self-learning tools, it says a client’s informed consent is required before the lawyer inputs information relating to the representation. It adds that “merely adding general, boiler-plate provisions to engagement letters purporting to authorize the lawyer to use GAI is not sufficient.” It also suggests reading a tool’s Terms of Use and privacy policy, or consulting someone who has, to understand who can see the data and how it is used.

When Does Opinion 512 Say to Disclose AI Use?

The opinion does not say every use of AI has to be disclosed to clients. Under Rule 1.4, it describes situations where disclosure is required: when a client asks, when the engagement terms or outside counsel guidelines call for it, when AI output will influence a significant decision in the representation, and when the use bears on the basis of the fee. Corporate legal departments that write outside counsel guidelines effectively get to set that trigger for the firms they hire.

On courts, the opinion links the duties of meritorious claims and candor, Rules 3.1, 3.3 and 8.4(c), to the problem of fabricated citations. It notes that “some courts have responded by requiring lawyers to disclose their use of GAI,” and points lawyers to local rules and standing orders. For managers, Rules 5.1 and 5.3 cover supervision: the opinion describes clear policies on permissible use, training, and due diligence on vendors, including their security, data retention and conflicts practices.

What Does Opinion 512 Say About Fees?

The fees section under Rule 1.5 gets the most attention inside firms. For hourly billing, the opinion says lawyers may bill only for time actually spent, and gives the example of a tool that drafts a pleading in 15 minutes: the lawyer bills the 15 minutes plus review time, not the hours the task used to take. Flat and contingent fees remain subject to the reasonableness test, and the opinion says that if a tool lets a lawyer finish tasks much more quickly, it may be unreasonable to charge the same flat fee as without it. Tools built into ordinary software, like a word processor, are treated as overhead, while a third-party tool used for a specific matter may be billed as an expense at cost. Under the opinion, a lawyer may not bill a client for learning a tool the lawyer will use regularly, though billing to learn a specific tool the client explicitly requests may be appropriate.

State guidance has built on this since. North Carolina’s 2024 Formal Ethics Opinion 1 reaches similar conclusions on time-based billing. Virginia took a different view of value pricing: LEO 1901, approved by the Supreme Court of Virginia on November 24, 2025, says it is “not per se unreasonable” to charge the same non-hourly fee for work done with AI as without it, while agreeing that hourly fees can only reflect actual time. Read side by side with 512’s flat-fee passage, that is a real difference in emphasis. Florida’s Bar has invited comments on a new opinion covering agentic AI and AI recording tools, which are areas Opinion 512 barely reaches.

What Should You Watch Next?

Opinion 512 was written before AI agents that carry out multistep work became common, and before AI notetakers began recording client calls. State bars are filling those gaps first, with New York City, Oregon and Florida all issuing or drafting guidance on recording and agents in the past year. Watch whether the ABA committee returns to the subject with an opinion on agents, and watch how often courts and disciplinary bodies cite 512 in sanctions and grievance decisions, since that is the real test of how much weight an advisory opinion carries.

FAQ

When was ABA Formal Opinion 512 issued?

July 29, 2024, per the ABA.

Does Opinion 512 require disclosing every AI use to clients?

No. It describes specific situations where disclosure is required, such as when a client asks or engagement terms call for it.

Can lawyers bill hourly for time AI saved?

No. The opinion says hourly billing covers only time actually spent.

Is Opinion 512 binding?

No. ABA opinions interpret the Model Rules, which states adopt in their own versions.