AI Regulation Tracker / Attorneys
The ABA's First Formal Ethics Opinion on Lawyers Using Generative AI
Formal Opinion 512, issued July 29, 2024, does not create new rules. It reads the existing Model Rules of Professional Conduct onto a lawyer's use of generative AI, and it is the national interpretive baseline that state bars across the country now cite.
The American Bar Association does not usually make headlines, and Formal Opinion 512 is not a statute, a rule, or a court order. It is an advisory ethics opinion. But it is the advisory ethics opinion on lawyers and generative AI, the first one the ABA has issued, and that is why it matters far beyond its formal weight. When a state bar in Florida or Texas or New York sits down to decide whether a lawyer crossed a line by using a chatbot, this is the document it opens first. Issued on July 29, 2024 by the ABA Standing Committee on Ethics and Professional Responsibility, it takes the Model Rules of Professional Conduct that already govern every lawyer and asks a narrow, practical question: what do these longstanding duties require when the tool in your hand is a large language model?
The committee is careful about its own reach, and so am I. The opinion creates no new obligation. It interprets duties that were already there. But the interpretation is authoritative in the sense that carries real consequences, because state disciplinary bodies overwhelmingly model their own rules on the ABA's and reach for ABA opinions when they apply them. Treating Opinion 512 as merely academic because it is "only" advisory is a mistake a lawyer makes once.
Competence means you understand the tool and check its work
The first duty is competence under Model Rule 1.1. The opinion's reassurance is that "lawyers need not become GAI experts." Its demand is that they understand, in its words, "the capabilities and limitations" of the specific tool they use. That includes the well-documented tendency of these systems to fabricate. The practical consequence is verification, and the opinion ties the intensity of that verification to the task. As it puts it, "the required verification is factually specific and will necessarily depend on the GAI tool and the specific task" it performs. Using a model to brainstorm arguments is not the same as relying on it to produce case citations you will file. The closer the output gets to something a client or a court relies on, the more independent checking the rule expects. This is the duty behind every sanctions order you have read about lawyers filing hallucinated cases. Opinion 512 is where the profession said, in advance, that you cannot outsource your judgment to the machine.
Confidentiality and the informed-consent trap
The second duty is confidentiality under Model Rule 1.6, and this is where the opinion is most likely to change day-to-day behavior. Before a lawyer inputs information relating to a representation into a GAI tool, the opinion says they must evaluate the risk that the information will be disclosed to or accessed by people outside the firm. For self-learning tools that train on what users feed them, the opinion goes further and states that "a client's informed consent is required prior to inputting information relating to the representation into such a GAI tool." Two things about that are easy to get wrong. First, informed consent means the client actually understands why the tool is being used and what the risk is; the opinion is explicit that boilerplate language buried in an engagement letter does not do the job. Second, the trigger is the tool's behavior, not the lawyer's intent. If the product learns from your inputs, the confidential nature of the client's information does not protect you from the consent requirement, it is the reason for it.
Communication, candor, and supervision
Three more duties round out the framework. Communication under Model Rule 1.4 can require telling a client that you are using GAI in their matter, particularly where the use is material to the representation or the client would reasonably want to know. Candor toward the tribunal under Model Rule 3.3 is the courtroom backstop to the competence duty: a lawyer remains fully responsible for the accuracy of anything submitted to a court, regardless of which tool drafted it. And supervision under Model Rules 5.1 and 5.3 pushes the obligation up the org chart. The opinion is direct that managerial lawyers must establish clear policies on the firm's permissible use of GAI, and that supervisory lawyers must make reasonable efforts to ensure lawyers and nonlawyers comply. In a firm, "I did not know an associate was pasting client files into a chatbot" is not a defense the rule leaves open.
Fees: you cannot bill for time the machine saved
The last duty is reasonable fees under Model Rule 1.5, and it is the one with the sharpest edge. If a GAI tool lets you do in ten minutes what used to take an hour, you bill the ten minutes. The opinion states the principle plainly: "a fee charged for which little or no work was performed is an unreasonable fee." It also addresses the cost of getting up to speed. A lawyer generally may not bill a client for the time spent learning to use AI tools as a matter of general competence, though the calculus can differ when a client specifically asks for a particular tool on their matter. The efficiency gain belongs to the client, not to the timesheet.
What US lawyers should do now
None of this is new law, and that is exactly why it is easy to neglect. The right response is not panic, it is documentation and policy. Know the specific tools your firm uses and what they do with input data. Build a real informed-consent conversation into intake for any tool that learns from what you feed it, and do not rely on a line in the engagement letter to carry it. Verify AI output in proportion to how much a client or court will rely on it, and keep the verification trail. If you supervise anyone, write the GAI policy down and enforce it, because Rule 5.1 makes that your problem. And review your billing practice so you are charging for work performed, not for time the tool eliminated. Every state bar that has spoken since has built on this opinion. Aligning to it now is how you stay ahead of the discipline your competitors will learn about the hard way.
Questions professionals are asking
Is ABA Formal Opinion 512 binding on lawyers?
No. It is an advisory ethics opinion interpreting the ABA Model Rules of Professional Conduct. The Model Rules become binding only when a state adopts them. What gives Opinion 512 real force is that state bars and courts routinely follow ABA opinions when they apply their own professional-conduct rules, so it operates as the national baseline even though it is not itself law.
Do I need a client's consent to use AI on their matter?
Often, yes. The opinion says a client's informed consent is required before inputting information relating to the representation into a self-learning GAI tool that trains on user inputs. Informed consent means the client actually understands why the tool is used and what the risk is. The opinion states that boilerplate language in an engagement letter is not enough.
Can I bill a client for time an AI tool saved me?
No. Under the reasonable-fee duty, you bill for work actually performed. The opinion states that a fee charged for which little or no work was performed is an unreasonable fee. If a tool lets you complete a task faster, the client gets the efficiency. Generally you also cannot bill a client for time spent learning AI tools as a matter of general competence.
Am I responsible if the AI invents a case citation?
Yes. Competence under Model Rule 1.1 requires you to understand a tool's limitations, including its tendency to fabricate, and to verify its output. Candor toward the tribunal under Rule 3.3 makes you fully accountable for anything submitted to a court, regardless of which tool drafted it. The required verification scales with how much a client or court will rely on the output.
What do firm leaders have to do under the opinion?
Set policy and supervise. Under Model Rules 5.1 and 5.3, managerial lawyers must establish clear policies on the firm's permissible use of generative AI, and supervisory lawyers must make reasonable efforts to ensure lawyers and nonlawyer staff comply. A written, enforced GAI policy is the practical expression of that duty.
RELATED BRIEFINGS
Browse the full AI Regulation News tracker
Informational analysis for working professionals, not legal advice. Confirm how any rule or ethics opinion applies to your situation with qualified counsel and the bar in your jurisdiction.