ABA Formal Opinion 512 on Generative AI: What It Actually Says, Rule by Rule (2026)

Last reviewed: October 2026. Quotations are verbatim; page numbers are the ABA page numbers in the opinion's footer. Primary source: ABA Formal Opinion 512 (official PDF).

Opinion 512 is cited constantly and paraphrased loosely. Some say it merely "recommends" consent, requires a written AI policy, or was updated for agentic AI. The text says none of those things. This reference gives each section's key sentence verbatim with its page, a plain-English reading, and what it means in practice.

The opinion describes its subject as "a subset of AI technology that has more recently drawn the attention of the legal profession and the world at large — generative AI (GAI)" (p. 1), and its own purpose modestly: "This Opinion identifies some ethical issues involving the use of GAI tools and offers general guidance for lawyers attempting to navigate this emerging landscape" (p. 2). Your own jurisdiction's rules are what bind you.

Opinion 512 at a Glance

A. Competence (Model Rule 1.1), pp. 2–5

To competently use a GAI tool in a client representation, lawyers need not become GAI experts. Rather, lawyers must have a reasonable understanding of the capabilities and limitations of the specific GAI technology that the lawyer might use. (pp. 2–3)

The standard is knowledge of the specific tool. Lawyers may acquire it themselves or "draw on the expertise of others" (p. 3), and it does not stand still: "This is not a static undertaking" (p. 3).

The verification standard

The opinion warns that some tools are prone to "hallucinations," which it describes as "providing ostensibly plausible responses that have no basis in fact or reality" (p. 3). Its verification sentence:

a lawyer's reliance on, or submission of, a GAI tool's output--without an appropriate degree of independent verification or review of its output--could violate the duty to provide competent representation as required by Model Rule 1.1. (pp. 3–4)

The key words are "appropriate degree". The review needed "will necessarily depend on the GAI tool and the specific task that it performs as part of the lawyer's representation of a client" (p. 4). In the opinion's contract-summary example, the lawyer "would not necessarily have to manually review the entire set of documents to verify the results if the lawyer had previously tested the accuracy of the tool on a smaller subset of documents" (p. 4).

Two limits hold regardless. Lawyers "may not abdicate their responsibilities by relying solely on a GAI tool to perform tasks that call for the exercise of professional judgment" (p. 4). And: "In short, regardless of the level of review the lawyer selects, the lawyer is fully responsible for the work on behalf of the client" (p. 4).

The "17% and 33%" hallucination figure in footnote 14 comes from a Stanford study (Magesh et al., June 26, 2024), not from the ABA. And there is no present duty to use AI, though "it is conceivable that lawyers will eventually have to use them" (p. 5).

What this means in practice

• Learn the specific tool you use, or rely on someone who knows it, and keep that current.
• Match review to the task; test a tool on a sample you can check before trusting it at scale.
• Keep legal advice and negotiation with a lawyer. Whatever you check or skip, the work is yours.

B. Confidentiality (Model Rule 1.6), pp. 6–7

Accordingly, 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. (p. 7)

The section rests on Rule 1.6(c), with similar protection for former and prospective clients under Rules 1.9(c) and 1.18(b) (p. 6). The first step is a risk assessment: "Before lawyers input information relating to the representation of a client into a GAI tool, they must evaluate the risks that the information will be disclosed to or accessed by others outside the firm" (p. 6). The analysis "will be fact-driven and depend on the client, the matter, the task, and the GAI tool used to perform it" (p. 6).

Self-learning tools: consent "is required"

The opinion's definition: "Some GAI tools are described as "self-learning," meaning they will learn from themselves as they cull more data" (p. 1). The risk exists "even if the tool is used exclusively by lawyers at the same firm" (p. 6). The wording on p. 7 is mandatory, but the trigger is narrow: inputting information relating to the representation into such a tool. Footnote 34 adds that the conclusion "is based on the risks and capabilities of GAI tools as of the publication of this opinion" (p. 7, fn 34).

What informed consent has to contain

For the consent to be informed, the client must 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. (p. 7)

And for engagement letters: "To obtain informed consent when using a GAI tool, merely adding general, boiler-plate provisions to engagement letters purporting to authorize the lawyer to use GAI is not sufficient" (p. 7).

Terms of use, and the idea-generation exception

"As a baseline, all lawyers should read and understand the Terms of Use, privacy policy, and related contractual terms and policies of any GAI tool they use to learn who has access to the information that the lawyer inputs into the tool or consult with a colleague or external expert who has read and analyzed those terms and policies" (p. 7). Lawyers "may need to consult with IT professionals or cyber security experts" (p. 7).

Consent is not always needed: "if a lawyer is using the tool for idea generation in a manner that does not require inputting information relating to the representation, client informed consent would not be necessary" (p. 7).

What this means in practice

• Before client information goes into a tool, learn from its terms who can access your inputs.
• For a self-learning tool, get matter-specific informed consent covering the p. 7 elements first. A generic AI clause is not that consent.
• Abstract brainstorming with no client information needs no consent under this section.

C. Communication (Model Rule 1.4), pp. 8–9

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. (p. 8)

Rule 1.4 is a separate inquiry from Rule 1.6 (p. 8). Disclosure is required in defined cases:

• When asked, or when terms require it. "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" (p. 8).
• Inputting client information, which needs advance notice and informed consent (p. 8).
• Fees. "Lawyers must also consult clients when the use of a GAI tool is relevant to the basis or reasonableness of a lawyer's fee" (p. 8).
• Significant decisions, "such as when a lawyer relies on GAI technology to evaluate potential litigation outcomes or jury selection" (pp. 8–9).
• Engagement terms or the client's reasonable expectations about how the work will be done (p. 9).

For everything else, "It is not possible to catalogue every situation in which lawyers must inform clients about their use of GAI." The factors: "the GAI tool's importance to a particular task, the significance of that task to the overall representation, how the GAI tool will process the client's information, and the extent to which knowledge of the lawyer's use of the GAI tool would affect the client's evaluation of or confidence in the lawyer's work" (p. 9).

Voluntary disclosure is always open: "The engagement agreement is a logical place to make such disclosures and to identify any client instructions on the use of GAI in the representation" (p. 9). Read this with Section B: the engagement letter is a good place to disclose, but boiler-plate there is not informed consent.

What this means in practice: answer honestly when asked, check outside counsel guidelines, and raise AI use yourself when it bears on the fee, a significant decision or the client's expectations.

D. Meritorious Claims and Candor (Model Rules 3.1, 3.3, 8.4(c)), pp. 9–10

Even an unintentional misstatement to a court can involve a misrepresentation under Rule 8.4(c). Therefore, output from a GAI tool must be carefully reviewed to ensure that the assertions made to the court are not false. (p. 10)

The problems seen so far: "citations to nonexistent opinions, inaccurate analysis of authority, and use of misleading arguments" (p. 10). Before submitting anything to a court, lawyers must review outputs "including analysis and citations to authority, and to correct errors, including misstatements of law and fact, a failure to include controlling legal authority, and misleading arguments" (p. 10). The opinion notes that "Some courts have responded by requiring lawyers to disclose their use of GAI" (p. 10), and footnote 47 tells lawyers to "consult with the applicable court's local rules" (p. 10, fn 47).

What this means in practice

• Read every authority a tool cites before filing, and look for controlling authority it left out.
• Check each court's local rules on AI disclosure. See our guide on avoiding AI hallucinations and sanctions.

E. Supervisory Responsibilities (Model Rules 5.1, 5.3), pp. 10–11

Managerial lawyers must establish clear policies regarding the law firm's permissible use of GAI, and supervisory lawyers must make reasonable efforts to ensure that the firm's lawyers and nonlawyers comply with their professional obligations when using GAI tools. (p. 10)

The duty is "clear policies". The opinion does not say the policy must be written. A written policy is good practice, not the text of 512.

Training. "Training could include the basics of GAI technology, the capabilities and limitations of the tools, ethical issues in use of GAI and best practices for secure data handling, privacy, and confidentiality" (p. 10). Footnote 52 adds a suggestion: "One training suggestion is that all materials produced by GAI tools be marked as such when stored in any client or firm file so future users understand potential fallibility of the work" (p. 10, fn 52).

Vendors. Under Rule 5.3, the opinion's due-diligence list for outside vendors (reference checks, security policies, confidentiality agreements, conflicts screening and others) carries over: "These concepts also apply to GAI providers and tools" (p. 11).

What this means in practice: decide which uses and tools are permitted and with what review; train people on them; consider labeling AI material in files; vet vendors before adoption. Our law firm AI policy template is one starting point. It is our template, not the ABA's.

F. Fees (Model Rule 1.5), pp. 11–14

GAI tools may provide lawyers with a faster and more efficient way to render legal services to their clients, but lawyers who bill clients an hourly rate for time spent on a matter must bill for their actual time. (p. 12)

Before charging for AI at all, "the lawyer must explain the basis for the charge, preferably in writing" (p. 12).

Hourly: the 15-minute example

"If a lawyer uses a GAI tool to draft a pleading and expends 15 minutes to input the relevant information into the GAI program, the lawyer may charge for the 15 minutes as well as for the time the lawyer expends to review the resulting draft for accuracy and completeness" (p. 12).

Flat and contingent fees

"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" (p. 12). The word is "may", not "must". Virginia has since taken the opposite view (see below).

Overhead, expenses and in-house tools

A tool that "functions similarly to equipping and maintaining a legal practice" is overhead, not charged "absent a contrary disclosure to the client in advance"; the example is a GAI grammar checker in word processing software (p. 13). A third-party GAI service charged per use to review "thousands of voluminous contracts for a particular client" may ordinarily be billed "as an expense for the actual out-of-pocket expense" (p. 13), and "the lawyer should not add a surcharge to the actual cost of such expenses and should pass along to the client any discounts" (p. 13). For a firm's own tool, absent agreed rates, it "may charge the client no more than the direct cost associated with the tool (if any) plus a reasonable allocation of expenses directly associated with providing the GAI tool" (p. 14).

Time spent learning a tool

a lawyer may not charge a client to learn about how to use a GAI tool or service that the lawyer will regularly use for clients because lawyers must maintain competence in the tools they use, including but not limited to GAI technology. (p. 14)

The exception: "if a client explicitly requests that a specific GAI tool be used in furtherance of the matter and the lawyer is not knowledgeable in using that tool, it may be appropriate for the lawyer to bill the client to gain the knowledge to use the tool effectively," with billing terms agreed beforehand and, preferably, memorialized (p. 14).

What this means in practice

• Hourly: bill input and review time, nothing for time the tool saved.
• Flat fees: if AI makes the work much faster, ask whether the same fee is still reasonable.
• Embedded tools are overhead; per-use costs for a client pass through at cost.
• No billing for learning tools you use regularly, unless the client asked for a specific tool and you agreed terms.

What Opinion 512 Does Not Say

Claims seen online, checked against the text:

• "It merely recommends informed consent." No. For self-learning tools, consent "is required" before inputting information relating to the representation (p. 7).
• "It requires consent before using AI at all." No. Idea generation without client information needs no consent (p. 7).
• "It requires disclosing every use of AI." No. "Depending on the circumstances, client disclosure may be unnecessary" (p. 8).
• "It requires a written AI policy." No. It requires "clear policies" (p. 10). "In writing" appears only for explaining the basis of a fee, "preferably in writing" (p. 12).
• "Boilerplate waivers will not suffice as informed consent." Right idea, wrong words. The text: "merely adding general, boiler-plate provisions to engagement letters purporting to authorize the lawyer to use GAI is not sufficient" (p. 7).
• "It approves or bans tools, or mandates security measures." No. The analysis is "fact-driven" (p. 6); footnote 31 cites Opinion 477R's fact-specific approach to security.
• "Lawyers can never bill for learning AI." The bar covers tools used regularly, with an exception for a client-requested tool (p. 14).
• "Flat fees must be cut when AI saves time." The text says the same fee "may be unreasonable" (p. 12).
• "The ABA found AI hallucinates 17 to 33% of the time." That is the Stanford study in footnote 14.
• "It addresses privilege waiver." No. "Privilege" appears once, describing e-discovery tools used "to segregate privileged documents" (p. 1). See our analysis of the Heppner ruling on AI and privilege.
• "It covers conflicts and advertising." No. Footnote 4 says issues under Rules 7.1, 1.7 and 1.9 "may surface" (p. 2, fn 4); footnote 47 flags Rule 3.4(c) (p. 10). None is analyzed.
• "It was updated in May 2026 for agentic AI." We found no source for this and no later ABA formal opinion revising 512.
• Six or seven duties? The headnote lists seven; the Discussion has six sections, A to F.

What Has Happened Since July 2024

The opinion expected follow-up. It calls GAI tools "a moving target--indeed, a rapidly moving target" and anticipates "updated guidance on professional conduct issues relevant to specific GAI tools as they develop" (p. 2).

ABA Formal Opinion 517 (July 9, 2025) addresses discrimination in jury selection, including reliance on AI tools. We could not confirm from its text whether 517 cites 512.

ABA Task Force on Law and Artificial Intelligence. Its report "Addressing the Legal Challenges of AI: Year 2 Report on the Impact of AI on the Practice of Law" was released on December 15, 2025, according to LawNext, which describes it as the Task Force's final report. It is not an ethics opinion.

State bars. 512's own footnotes cite earlier guidance, including Florida Bar Opinion 24-1, Pennsylvania and Philadelphia Joint Formal Opinion 2024-200, the State Bar of California's practical guidance, West Virginia Opinion 24-01 and DC Bar Opinion 388. Selected documents:

Virginia takes the opposite view on flat fees. LEO 1901 states: "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," and "Rule 1.5 does not equate reduced time with proportionally reduced fees." Its March 20, 2025 draft named Opinion 512 as a contrary view; the approved text does not name it.

Texas 705 summarizes the ABA and Florida opinions as suggesting "that lawyers should obtain informed consent before using these tools", broader than 512's actual trigger.

California. The State Bar's AI page lists 512 as a resource and lists "Proposed Amendments to the Rules of Professional Conduct Related to Artificial Intelligence" under 2026 public comment. We did not find their outcome.

Status. As of October 2026, no later ABA formal opinion revising 512 was found. For wider state coverage, see our overview of AI legal ethics and bar association guidelines.

An Opinion 512 Compliance Checklist

1. Know each tool's capabilities and limits, or rely on someone who does (pp. 2–3).
2. Set a review level per tool and task; keep judgment calls with a lawyer (p. 4).
3. Read each tool's Terms of Use and privacy policy before inputting client information, and assess disclosure risk (pp. 6–7).
4. Get informed consent with the p. 7 elements before inputting client information into a self-learning tool; no boiler-plate (p. 7).
5. Disclose AI use when asked, when terms require it, and when it bears on fees or significant decisions (pp. 8–9).
6. Verify every citation and check court local rules before filing (p. 10, fn 47).
7. Adopt clear firm policies; train lawyers and nonlawyers (p. 10).
8. Run vendor due diligence on GAI providers (p. 11).
9. Explain any AI charge in advance, preferably in writing; bill actual time (p. 12).
10. Treat embedded tools as overhead; pass per-use costs through at cost (p. 13).
11. Do not bill for learning tools you use regularly (p. 14).

Applying Opinion 512 When You Use AI Prompts

We sell AI prompt libraries and document generators for lawyers, so here is how 512 applies to tools like ours. No tool makes a lawyer compliant; the duties belong to the lawyer.

• Verification (p. 4). Our prompts are written to flag missing authority rather than invent it, and our generators include anti-hallucination rules. On the Strategic plan, generators show a visible Reasoning Log so you can see how a draft was built. This can make review faster. It does not replace it.
• Confidentiality (p. 7). The baseline applies to us like any vendor: read the terms and privacy policy before putting client information into any tool.
• Idea generation (p. 7). Many prompts work for structure and issue-spotting without client information.

You can browse the prompt library to see how the prompts are built.

Related Reading

• Law Firm AI Policy Template: our written policy template organized around 512's duties (ours, not the ABA's).
• AI Legal Ethics and Bar Association Guidelines: the broader ethics picture and state guidance.
• AI Hallucinations in Legal Work: a method for the review duty in Sections A and D.
• AI and Attorney-Client Privilege After Heppner: privilege is a separate question 512 does not address.

This article summarizes ABA Formal Opinion 512 for general information. It is not legal advice. ABA formal opinions are not binding on their own; the rules and ethics opinions of your jurisdiction govern your conduct. Read the full opinion and consult your state bar or ethics counsel on specific questions.