What does ABA Formal Opinion 512 say about telling clients you used AI?
ABA Formal Opinion 512, issued 29 July 2024 by the ABA Standing Committee on Ethics and Professional Responsibility under the title “Generative Artificial Intelligence Tools,” treats disclosure under Model Rule 1.4 as fact-dependent: “The facts of each case will determine whether Model Rule 1.4 requires lawyers to disclose their GAI practices to clients or obtain their informed consent to use a particular GAI tool. Depending on the circumstances, client disclosure may be unnecessary.”
It then names the situations that are not optional. Lawyers “must disclose their GAI practices if asked by a client how they conducted their work, or whether GAI technologies were employed in doing so, or if the client expressly requires disclosure under the terms of the engagement agreement or the client’s outside counsel guidelines.” Clients “would need to be informed in advance, and to give informed consent, if the lawyer proposes to input information relating to the representation into the GAI tool.” Lawyers “must also consult clients when the use of a GAI tool is relevant to the basis or reasonableness of a lawyer’s fee.” The opinion adds that consultation is necessary when the output “will influence a significant decision in the representation.”
The structural point is easy to miss and worth stating plainly: the opinion reaches its firmest conclusions under Model Rule 1.6, not Model Rule 1.4. The question that actually forces a conversation with the client is usually whether client information is going into the tool, not whether a tool was used.
When is client consent required rather than merely a good idea?
Under ABA Formal Opinion 512, consent is required whenever information relating to the representation is going into a self-learning tool: “because many of today’s self-learning GAI tools are designed so that their output could lead directly or indirectly to the disclosure of information relating to the representation of a client, a client’s informed consent is required prior to inputting information relating to the representation into such a GAI tool.” The opinion attaches a footnote saying that conclusion is based on the risks and capabilities of these tools as of its publication and may change as the technology does.
Informed is doing real work in that sentence. The opinion requires the client to have “the lawyer’s best judgment about why the GAI tool is being used, the extent of and specific information about the risk, including particulars about the kinds of client information that will be disclosed, the ways in which others might use the information against the client’s interests, and a clear explanation of the GAI tool’s benefits to the representation.” It then closes the obvious workaround: “merely adding general, boiler-plate provisions to engagement letters purporting to authorize the lawyer to use GAI is not sufficient.”
It also marks where consent is not required. Where the lawyer will not be inputting information relating to the representation — the opinion’s example is idea generation — client informed consent is not necessary. The New York City Bar Association Committee on Professional Ethics reached a compatible position in Formal Opinion 2024-5 (7 August 2024), which states that a lawyer must not input confidential client information into any generative AI system lacking adequate confidentiality and security protections without informed consent, and that a lawyer “should consider disclosing” the intent to use generative AI that is not generally understood to be routinely used by lawyers. Both are advisory, and the New York City Bar opinion speaks to New York lawyers. Which deployment tier changes that analysis is covered in which AI tier actually protects client data.
Where have state bars gone further than the ABA, or in a different direction?
Both directions are represented, which is why a national summary of this is not usable. Florida Bar Ethics Opinion 24-1, approved 19 January 2024, is softer than the ABA on the same question: it says that “it is recommended that a lawyer obtain the affected client’s informed consent prior to utilizing a third-party generative AI program if the utilization would involve the disclosure of any confidential information,” under Rule 4-1.6 of the Rules Regulating The Florida Bar. Recommended is not required, and the Florida opinion states on its face that advisory ethics opinions are not binding.
North Carolina 2024 Formal Ethics Opinion 1, adopted 1 November 2024, goes further than the ABA in a different place. It distinguishes what it calls ordinary tasks, such as conducting legal research or generic case and practice management, from substantive delegation, and holds that where “a lawyer delegates substantive tasks in furtherance of the representation to an AI tool, the lawyer’s use of the tool is akin to outsourcing legal work to a nonlawyer or other third-party resource or service, for which the client’s advanced informed consent is required.” That is a consent trigger keyed to what the tool is doing rather than to what data goes in.
The District of Columbia Bar issued Ethics Opinion 388, “Attorneys’ Use of Generative Artificial Intelligence in Client Matters,” in April 2024, framing the inquiry around whether information given to the tool will be visible to third parties and whether it will affect answers given to future users, without imposing blanket disclosure. Every one of these is advisory and jurisdiction-specific. A firm admitted in several states is reading several of them, and in practice the strictest tends to set the operating standard. The Institute’s bar guidance, state by state area tracks the divergence rather than averaging it.
What are outside counsel guidelines starting to require?
Whatever they require, they bind by contract rather than by rule, and ABA Formal Opinion 512 treats them as a disclosure trigger in their own right: a lawyer must disclose GAI practices where “the client expressly requires disclosure under the terms of the engagement agreement or the client’s outside counsel guidelines.” New York City Bar Formal Opinion 2024-5 makes the corresponding housekeeping point, that a lawyer “should review any applicable client instructions or guidelines that may restrict or limit the use of Generative AI.”
What those guidelines actually say is largely unpublished, because they are private contracts between a client and a firm, so no reliable public tally exists. What can be dated is the visibility gap. The Thomson Reuters Institute reported on 18 March 2026, from its 2026 AI in Professional Services Report, that more than half of corporate legal professionals surveyed thought their outside firms should be using AI on their matters, while 68 percent said they had no idea whether their outside firms were using it.
Read that as an operational warning rather than a statistic. A firm with no single answer to the question “how do we use AI on your matters” will give several different answers across several partners the first time a general counsel asks. Deciding the position once and writing it down is one of the five decisions in the firm AI policy.
How does billing for AI-assisted work interact with the duty of reasonable fees?
On hourly billing there is no disagreement. ABA Formal Opinion 512 applies ABA Formal Opinion 93-379 and states that lawyers “who bill clients an hourly rate for time spent on a matter must bill for their actual time.” Its worked example is deliberately small: a lawyer who spends fifteen minutes inputting information to draft a pleading may charge for those fifteen minutes and for the time spent reviewing the draft for accuracy and completeness. North Carolina 2024 Formal Ethics Opinion 1 reaches the same place, answering no to a lawyer who wanted to bill three hours for estate planning documents that took one hour with AI.
Non-hourly fees are where the two documents pull in different directions. ABA Formal Opinion 512 says that “if using a GAI tool enables a lawyer to complete tasks much more quickly than without the tool, it may be unreasonable under Rule 1.5 for the lawyer to charge the same flat fee when using the GAI tool as when not using it.” Virginia Legal Ethics Opinion 1901, approved by the Supreme Court of Virginia on 24 November 2025, starts from the other end: “It is not per se unreasonable for a lawyer to charge the same non-hourly fee for work done with the assistance of AI as work done without the use of AI,” reasoning that the “skill requisite to perform the legal service properly” may rise even as time falls, and that Rule 1.5 “does not equate reduced time with proportionally reduced fees.” The two are not flatly contradictory — the ABA says the same flat fee may be unreasonable on particular facts, Virginia says it is not unreasonable as a matter of course — but Virginia adds that “[c]ontrary views fail to appreciate the value of advancing technology,” without naming the ABA opinion.
Cost is a third question again. ABA Formal Opinion 512 treats a tool that functions like equipping and maintaining a practice — its example is a grammar checker in word processing software — as overhead that should not be charged absent contrary disclosure in advance, while a third-party service billed per use to review a particular client’s contracts would ordinarily be billable as an actual out-of-pocket expense. Virginia LEO 1901 adds a communication duty under Rule 1.5(b): where time spent is substantially reduced, the client may need additional explanation of why the lawyer’s experience and skill still support the fee. What this does to the economics is covered in what AI does to the billable hour and in billing AI-assisted work.
What is the difference between a rule, an opinion and a practice here?
A rule binds, an opinion interprets, and a practice is just what firms do — and almost everything described above is in the second and third categories. Rules of professional conduct are adopted state by state and are the only instruments that bind. ABA Formal Opinion 512 interprets the ABA Model Rules, which no lawyer is governed by directly, because the ABA does not regulate lawyers. Florida Bar Ethics Opinion 24-1 says on its own first line that advisory ethics opinions are not binding, and the same is true of the North Carolina, New York City and District of Columbia opinions in their own jurisdictions.
Virginia Legal Ethics Opinion 1901 sits differently, having been approved by the Supreme Court of Virginia on 24 November 2025 rather than issued by a bar committee alone. That is a higher order of endorsement than most of these documents carry, and it is still a Virginia document about Virginia’s Rule 1.5.
The practice layer is separate again: engagement letters, outside counsel guidelines and a firm’s own disclosure position are choices until a client makes one of them a term. This post is not legal advice and not ethics advice, and none of it states what any jurisdiction requires; that is a question for the firm’s own professional-responsibility counsel in the relevant state. The disclosure conversation itself is in client disclosure, and what clients have done about it is in are clients leaving firms over AI.