# ABA Formal Opinion 512 Explained, Duty by Duty

Published and updated 2026-10-29 · RuleX editorial team · United States · American English

ABA Formal Opinion 512, issued July 29, 2024, applies the Model Rules to generative AI: competence (1.1), confidentiality (1.6), communication (1.4), candor and meritorious claims (3.1, 3.3, 8.4(c)), supervision (5.1, 5.3) and reasonable fees (1.5). In practice, verify every output, obtain informed client consent before confidential information enters a self-learning tool and bill actual time. RuleX opens each source from the answer, so review happens inside the matter.

## Opinion 512 applies seven existing duties to every generative AI tool.

The ABA Standing Committee on Ethics and Professional Responsibility issued Formal Opinion 512 on July 29, 2024. It addresses generative AI tools that create new text, images, audio, video or code in response to a prompt, and it covers general-purpose products and tools designed for the practice of law alike. The opinion opens by naming the duties it applies: competent representation, protection of client information, communication with clients, supervision of employees and agents, advancing only meritorious claims and contentions, candor toward the tribunal and reasonable fees.

Read the opinion as a map of rules you already know. It rests on the ABA Model Rules as amended through August 2023, and it expects state and local bar ethics committees to add guidance of their own. The rules that govern your license are the rules of your jurisdiction, so use Opinion 512 to organize the analysis, then confirm each point against your state’s version of the rule and any state opinion. The state-by-state guide in this series collects the opinions verified against official sources.

The opinion also supplies the frame for everything that follows. The tools change quickly, the guidance is general, and the lawyer is fully responsible for the work on behalf of the client regardless of the level of review the lawyer selects.

## Competence (Rule 1.1): understand the tool, then verify what it produces.

Rule 1.1 requires the legal knowledge, skill, thoroughness and preparation reasonably necessary for the representation, and Comment [8] extends that duty to the benefits and risks of the technology you use. Opinion 512 sets a practical bar. Lawyers need not become experts in generative AI, but they need a reasonable understanding of what the specific tool does and where its output needs checking, or they must draw on someone who has that understanding. The task continues over time because the tools evolve; the opinion suggests reading about legal-profession tools, attending CLE programs and consulting colleagues who are proficient.

Verification scales with the tool and the task. The opinion’s own example is a lawyer who uses a tool to summarize numerous lengthy contracts. That lawyer does not necessarily have to re-read the entire set if the lawyer first tested the tool on a smaller subset, compared its summaries with a manual review and found them accurate. A tool designed for legal practice or for a discrete task, such as generating ideas, may call for less verification where the lawyer’s prior experience gives a reasonable basis to rely on the results. Submitting output without an appropriate degree of independent verification or review could violate Rule 1.1.

Judgment stays with the lawyer. The opinion states that lawyers may not leave it to a generative AI tool alone to offer legal advice to clients, negotiate their claims or perform other functions that require the lawyer’s personal judgment or participation. It adds that lawyers should become aware of the tools relevant to their work so they can decide, as a matter of professional judgment, whether to use them, and that the decision must account for the client’s interests and objectives.

Run the opinion’s own test on a closed matter. Upload up to twenty documents to RuleX, ask for the summary or chronology you would normally prepare, open the sources the answer cites and compare the result with your manual work. Keep that comparison with your firm’s AI-use file. It documents the reasonable understanding Rule 1.1 asks you to have.

## Confidentiality (Rule 1.6): evaluate the risk before client information goes in.

Rule 1.6(c) requires reasonable efforts to prevent the inadvertent or unauthorized disclosure of, or unauthorized access to, information relating to the representation. Before inputting client information into a generative AI tool, Opinion 512 asks the lawyer to weigh the likelihood of disclosure and unauthorized access, the sensitivity of the information, the difficulty of implementing safeguards and the extent to which safeguards affect the lawyer’s ability to represent the client. The analysis covers access outside the firm and inside it, for example by lawyers who sit behind an ethical wall. The answer depends on the client, the matter, the task and the tool.

The opinion draws a firm line for self-learning tools. Because their output could lead directly or indirectly to the disclosure of one client’s information to other users, or to other lawyers in the firm, it concludes that a client’s informed consent is required before information relating to the representation is input into such a tool. Informed consent means the client receives the lawyer’s best judgment about why the tool is used, specific information about the risk and the kinds of client information involved, the ways others might use that information against the client’s interests and a clear explanation of the tool’s benefits. General boilerplate added to an engagement letter is not sufficient.

As a baseline, read the tool’s terms of use, privacy policy and related contractual terms, or consult a colleague or outside expert who has. Where no information relating to the representation is input, for example in general idea generation, client consent under Rule 1.6 is not required. Each of these steps is easier when the vendor publishes its terms plainly. RuleX publishes its terms and privacy policy for U.S. users, and the vendor-checklist guide in this series shows how to read them against Opinion 512.

## Communication (Rule 1.4): the facts of the matter decide when to tell the client.

Opinion 512 sets a fact-based standard for disclosure. It says the facts of each case determine whether Rule 1.4 requires a lawyer to disclose generative AI practices and that, depending on the circumstances, disclosure may be unnecessary. It then lists the situations in which the lawyer must speak: when a client asks how the work was done or whether AI was used; when the engagement agreement or outside counsel guidelines require disclosure; when Rule 1.6 requires informed consent before information is input; when the use of the tool bears on the basis or reasonableness of the fee; and when the output will influence a significant decision, such as using AI to evaluate potential litigation outcomes or select a jury.

A lawyer may also tell clients how AI assists in delivering legal services when no rule requires it, and the opinion identifies the engagement agreement as the logical place to disclose AI practices and record client instructions. California’s guidance adds that a lawyer must follow client instructions or guidelines that restrict the use of AI. A short engagement-letter paragraph, supplemented by matter-specific consent when confidential information will enter a self-learning tool, addresses the Rule 1.4 and Rule 1.6 points together.

## Candor, meritorious claims and supervision (Rules 3.1, 3.3, 5.1, 5.3): the signature and the policy.

For litigation, the opinion points to Rules 3.1, 3.3 and 8.4(c). Rule 3.3(a) bars knowingly making a false statement of fact or law to a tribunal or failing to correct one, and the opinion observes that even an unintentional misstatement to a court can involve a misrepresentation under Rule 8.4(c). It records that issues arising to date include citations to nonexistent opinions, inaccurate analysis of authority and misleading arguments. Its instruction is concrete: before submitting materials to a court, review the outputs, including analysis and citations to authority, and correct errors such as misstatements of law and fact, a failure to include controlling legal authority and misleading arguments. The opinion also tells lawyers to check the applicable court’s local rules, because some courts require disclosure of AI use. The court-rules guide in this series lists verified examples.

Rules 5.1 and 5.3 convert individual care into firm policy. Managerial lawyers must establish clear policies on the firm’s permissible use of generative AI, and supervisory lawyers must make reasonable efforts to ensure that lawyers and nonlawyers comply, including training on the tools, the ethical issues and secure data handling. The opinion suggests marking all materials produced by generative AI tools as such when they are stored in a client or firm file, so later users understand the work needs review.

For outside providers, Rule 5.3(b) supports the vendor diligence the opinion lists: confidentiality configured and enforceable, notice to the lawyer of a breach or of service of process seeking client information, review of the provider’s reliability, security measures and liability terms, and clarity on whether the provider retains information or asserts rights in it.

## Fees (Rule 1.5): charge for actual time and for reasonable, disclosed expenses.

Hourly clients pay for the time you actually spend. Opinion 512 gives a clear example: if a lawyer spends fifteen minutes inputting information into a tool to draft a pleading, the lawyer may bill those fifteen minutes plus the time spent reviewing the draft for accuracy and completeness. For flat and contingent fees, the Rule 1.5(a) factors still apply, and the opinion says charging the same flat fee for work a tool made much faster may be unreasonable. Tool costs that function like equipping and maintaining a practice are overhead absent advance disclosure, while per-use charges for a client’s matter can be billed at actual cost. The billing guide in this series applies these rules in detail.

Put the duties into one workflow: test the tool on a sample, decide whether the client’s informed consent is required, describe your AI practices in the engagement agreement, open and verify every authority before you sign, mark AI-assisted materials in the file and bill the real time. Start the 7-day free trial, add a real matter and open the sources from the answer.

## Frequently asked questions

### What does ABA Formal Opinion 512 require lawyers to do?

It applies existing Model Rules to generative AI. Understand the tool you use, verify its output before relying on it or filing it, protect client information with informed consent before confidential information enters a self-learning tool, communicate with clients where the facts call for it, supervise lawyers and staff under clear firm policies and charge reasonable fees based on actual time.

### Is Opinion 512 binding on my license?

Opinion 512 is the ABA Standing Committee on Ethics and Professional Responsibility’s interpretation of the Model Rules. The rules of professional conduct in your jurisdiction control, so treat the opinion as persuasive guidance and check your state bar’s own opinions on the same questions.

### Do I have to tell clients that I use generative AI?

The opinion says the facts decide. Tell the client when the client asks, when the engagement agreement or outside counsel guidelines require disclosure, when consent is needed before confidential information is input, when the tool bears on the fee and when the output will influence a significant decision. You may also describe your AI practices in the engagement agreement.

### When is informed consent required?

The opinion concludes that informed consent is required before information relating to a representation is input into a self-learning generative AI tool whose output could lead to disclosure of that information. The lawyer explains why the tool is used, the specific risks and the benefits to the representation; boilerplate in an engagement letter is insufficient.

### Can I bill a client for the time a generative AI tool saves?

Hourly billing reflects actual time, including the time to prompt the tool and to review its output. Under flat and contingent arrangements, the Rule 1.5(a) factors decide reasonableness, and the opinion says charging the same flat fee for much faster work may be unreasonable.

### How does RuleX fit the verification Opinion 512 describes?

RuleX opens each source from the answer, works with up to twenty documents while keeping citations and context, and produces an editable Word draft beside the matter. You review the authority, revise the draft and keep the record with the case.

## Sources

- [ABA Formal Opinion 512 — Generative Artificial Intelligence Tools (July 29, 2024)](https://www.americanbar.org/content/dam/aba/administrative/professional_responsibility/ethics-opinions/aba-formal-opinion-512.pdf)

- [ABA Model Rules of Professional Conduct — table of contents](https://www.americanbar.org/groups/professional_responsibility/publications/model_rules_of_professional_conduct/model_rules_of_professional_conduct_table_of_contents/)

- [State Bar of California — Practical Guidance for the Use of Generative AI in the Practice of Law (2026 edition)](https://www.calbar.ca.gov/Portals/0/documents/ethics/Generative-AI-Practical-Guidance.pdf)

## Work on the matter in RuleX

[Create a U.S. account](https://rulex.org/united-states?register=1) · [U.S. pricing](https://rulex.org/united-states/pricing)

## Representation

This Markdown represents the same public page. Browsers receive HTML at its canonical URL.

- **Canonical (HTML):** https://rulex.org/united-states/guides/aba-formal-opinion-512
- **This representation:** https://rulex.org/united-states/guides/aba-formal-opinion-512.md
- **Home:** https://rulex.org/united-states
- **llms.txt:** https://rulex.org/llms.txt