ABA Formal Opinion 512 Tells Solo Lawyers Exactly What to Do About AI. Here It Is, Without the Legalese.

The ABA's first formal AI ethics guidance isn't a ban, a warning, or a vague caution. It's a five-part checklist. Here's what it actually requires of you this week.

By John Margerison · Sat, 23 May 2026 · 2176 words

Picture this: it's a Tuesday afternoon. You've got a motion due Thursday, a client intake at four, and your paralegal just told you she used ChatGPT to draft the first pass on a contract. You didn't ask her to. You didn't tell her not to. You have no idea what she typed into the prompt box.

That moment , the one you haven't had yet but probably will , is exactly what ABA Formal Opinion 512 was written for.

On July 29, 2024, the ABA Standing Committee on Ethics and Professional Responsibility issued its first formal opinion on generative AI in legal practice. Fifteen pages. Five model rules. Zero ambiguity about the direction of travel. Most of the press coverage reduced it to "lawyers must be careful with AI," which is roughly as useful as telling a surgeon to "be careful with knives." The opinion is more specific than that. And for a solo practitioner or small-firm partner, the specifics matter.

Here is what Opinion 512 actually requires, rule by rule.

RULE 1.1 , COMPETENCE: YOU HAVE TO UNDERSTAND THE TOOL, NOT JUST USE IT

Model Rule 1.1 requires lawyers to provide competent representation, which includes the "legal knowledge, skill, thoroughness and preparation reasonably necessary for the representation." The rule's Comment 8 has long required lawyers to keep up with "the benefits and risks associated with relevant technology." Opinion 512 applies that comment directly to generative AI.

What does that mean in practice? It means you cannot ethically use a tool you don't understand well enough to catch its errors. The opinion is explicit that AI tools "may combine otherwise accurate information in unexpected ways to yield false or inaccurate results" and are prone to "hallucinations" , producing plausible-sounding but fabricated content. You already know this from the news. What the opinion adds is that knowing about hallucinations in the abstract is not enough. You must understand the specific tool you're using: how it was trained, what its known failure modes are, and whether its outputs require verification before you rely on them.

Ask yourself: if the AI tool you used last week produced a wrong answer, would you have caught it? If the honest answer is "maybe not," that's a competence gap the opinion requires you to close.

The California State Bar's November 2023 Practical Guidance, which Opinion 512 cites approvingly, adopted a "reasonable efforts standard" and a "fact-specific approach" to technological competence. That framing is useful for solos. You are not required to become an AI engineer. You are required to understand the tool well enough to supervise its output , to know what it gets wrong, how often, and in what contexts. For a solo doing family law in a state with a well-defined statutory framework, that might mean learning that your AI drafting tool has a training cutoff that predates a recent statutory amendment. For a solo doing federal immigration work, it means knowing that AI tools frequently confuse agency guidance with binding regulation.

The practical floor: read the tool's documentation, understand its data cutoff, and verify every citation it produces. Every single one.

RULE 1.6 , CONFIDENTIALITY: INFORMED CONSENT IS NOT A CHECKBOX

This is the rule that generates the most confusion, and Opinion 512 is the most specific here.

The duty of confidentiality under Rule 1.6 covers "all information relating to the representation of a client, regardless of its source." Before you input any such information into a generative AI tool, the opinion says you must "evaluate the risks that the information will be disclosed to or accessed by others outside the firm."

The opinion then draws a line that most lawyers have missed. It distinguishes between two categories of AI tools: those that are "self-learning" (meaning they train on your inputs and could surface that information in responses to other users) and those that are not. For self-learning tools, the opinion concludes that "a client's informed consent is required prior to inputting information relating to the representation into such a GAI tool." This is not a soft recommendation. It is a requirement.

Here is what informed consent actually looks like, per the opinion: it is not a boilerplate clause buried in your engagement letter. The ABA's own commentary on Opinion 512 states plainly that "boilerplate consent included in engagement letters will not be adequate." Informed consent under Rule 1.0(e) requires that the client be given "adequate information and explanation about the material risks of and reasonably available alternatives to the proposed course of conduct." In plain English: you have to tell the client, in terms they understand, that you are using a tool that may process their information, explain what that means for their data, and get their actual agreement.

What does a compliant disclosure look like for a solo? Something like this, added to your engagement letter and explained verbally at intake: "Our firm uses AI-assisted drafting and research tools. Some of these tools may process information you share with us. We take steps to use tools with strong data-security protections, but we want you to understand this practice and have your consent to it. If you have concerns, please tell us and we will discuss alternatives."

That is not a legal template. It is a floor. Your state bar may require more. The Florida Bar's Ethics Opinion 24-1, approved January 19, 2024, requires lawyers to "take reasonable precautions to protect the confidentiality of client information" and specifically calls for lawyers to investigate AI providers' "data retention, sharing, and self-learning policies." That means before you use a tool with client data, you need to read the terms of service , or at minimum, understand whether the tool offers a data-privacy mode or enterprise agreement that prevents your inputs from being used for training.

One practical shortcut: many AI tools now offer a "no training" or "private" mode. Enabling that mode does not eliminate your disclosure obligation, but it does reduce the risk profile that triggers the consent requirement in the first place.

RULE 1.4 , COMMUNICATION: WHEN DO YOU HAVE TO TELL THE CLIENT YOU USED AI?

Opinion 512 addresses this more carefully than most summaries suggest. The rule requires lawyers to keep clients "reasonably informed" and to "explain a matter to the extent reasonably necessary to permit the client to make informed decisions."

The opinion does not require you to disclose every AI-assisted task. If you use AI for internal idea generation, research organization, or administrative work that does not involve inputting client-specific information, no disclosure is required. The trigger is when AI use is material to the representation , particularly when it involves client data, significantly shapes the work product, or when the client would reasonably want to know.

The NYC Bar's Formal Opinion 2024-5 frames this well: the question is whether the client would consider the information significant in deciding how to proceed. A client who hired you for your judgment and expertise might reasonably want to know if the first draft of their contract was generated by an AI tool rather than written by you. That is a conversation worth having proactively, not defensively.

RULES 5.1 AND 5.3 , SUPERVISION: YOUR PARALEGAL'S AI USE IS YOUR PROBLEM

Go back to that Tuesday afternoon. Your paralegal used ChatGPT on a client contract. Under Rules 5.1 and 5.3, that is your ethical exposure, not hers.

Opinion 512 is direct: "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 nonlawyer staff comply with the Rules of Professional Conduct." For a solo with one paralegal, "managerial lawyer" and "supervisory lawyer" are both you.

The opinion treats AI tools used by nonlawyer staff the same way it treats any nonlawyer assistance , you are responsible for the work product and for ensuring the staff member's use of the tool does not violate client confidentiality or produce unreviewed output that goes to a client or court. The Florida Bar's Opinion 24-1 explicitly treats AI tools "similarly to non-lawyer assistants under Rule 4-5.3(a), requiring proper supervision."

What does supervision look like in a two-person office? It means having a written policy , even a one-page document , that specifies: which AI tools are approved for use, what client information may or may not be entered into those tools, and that all AI-generated work product must be reviewed and verified by the supervising attorney before it goes anywhere. It means your paralegal knows the policy, has signed off on it, and understands that "the AI said so" is not a defense.

This is the documented-policy expectation the opinion creates. It does not require a 40-page compliance manual. It requires something written, communicated, and followed.

RULE 1.5 , FEES: THE EFFICIENCY PROBLEM IS REAL

This is the rule that will cause the most friction as AI becomes standard practice, and Opinion 512 takes a clear position.

If you bill hourly, you bill actual time. The opinion states: "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." If AI allows you to draft a motion in 15 minutes that used to take two hours, you bill 15 minutes of drafting time plus your review time. Not two hours.

The opinion also addresses flat fees. If AI compresses the work so dramatically that a flat fee set under pre-AI assumptions is now wildly disproportionate to the time and effort involved, that fee may no longer be reasonable under Rule 1.5. The opinion says directly: "if using a [GenAI] 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 [GenAI] tool as when not using it."

Two more billing rules from Opinion 512 that solos often miss. First, you cannot bill a client for the time you spend learning how to use an AI tool you will use generally across matters , that is overhead, not billable time. The exception is narrow: if a client specifically requests use of a particular AI tool for their matter, you may charge for learning that specific tool. Second, if you pass through the cost of a proprietary AI tool as an expense, you must explain the basis for that charge to the client, preferably in writing, and get their consent.

The practical upshot: update your engagement letter to address AI costs. Decide now whether your AI subscription is overhead (most solos should treat it this way) or a pass-through, and document that decision.

THE DOCUMENTED-POLICY EXPECTATION: WHAT YOU NEED IN WRITING

Opinion 512 does not use the phrase "written policy" as a formal requirement, but the supervision obligations under Rules 5.1 and 5.3 functionally require one. If you are ever asked to demonstrate that you exercised "reasonable efforts" to ensure your firm's AI use complied with the rules, a written policy is your evidence.

For a solo or small firm, a compliant AI policy document needs to cover five things: (1) which tools are approved for use and for what tasks; (2) what client information may be entered into those tools and under what conditions; (3) the consent and disclosure process for clients; (4) the verification requirement for all AI-generated work product before it goes to a client or court; and (5) how AI time and costs are handled in billing. One page. Kept in your office manual. Updated when you add a new tool.

That document is not bureaucracy. It is your defense.

WHAT TO DO THIS WEEK

Opinion 512 is not a prohibition. It is a framework. The ABA's own summary notes that "it is too early to claim that GAI tools are required for competent representation" , but it is equally clear that using them without understanding them, without client consent where required, and without supervision policies in place is an ethical violation waiting to happen.

The lawyers who will have problems are not the ones using AI. They are the ones using it carelessly , letting staff run client data through consumer tools without consent, billing two hours for work that took twenty minutes, and having no written record of any of it.

You can close most of your exposure this week. Read the terms of service on every AI tool you use and determine whether it is self-learning. Add a plain-language AI disclosure to your engagement letter and get client consent before inputting their information. Write a one-page AI use policy for your office and review it with any staff. Audit your billing practices to make sure AI efficiency gains are reflected in your invoices. And pick one CLE course on AI in legal practice , your state bar almost certainly offers one , and complete it before year-end. That is not a heavy lift. It is what the rules now require.