ABA Formal Opinion 512, in practice.

In July 2024 the ABA Standing Committee on Ethics and Professional Responsibility issued Formal Opinion 512, the first formal ABA guidance on generative AI. It does not ban AI and it does not bless any particular tool. It does something more durable: it maps the duties you already have under the Model Rules onto AI use, and it makes the lawyer, not the vendor, responsible for the result.

The practical reading is simple. If AI touches a matter, the same obligations still apply, and you should be able to show how you met them.

Opinion 512 ties AI use to familiar professional duties. None of them is new. What is new is having to show how each one holds up once a tool enters the workflow.

Competence (Rule 1.1)

You need a reasonable understanding of a tool's capabilities and limits as they relate to the matter. That does not mean becoming an engineer. It means knowing when output has to be verified rather than trusted, and not relying on a tool you cannot evaluate.

Confidentiality (Rule 1.6)

Client information cannot flow into a tool or vendor pipeline that is not safe. Self-learning systems that train on their inputs are a particular concern. Before client data goes in, you should know the vendor's data posture and, where needed, have the client's informed consent.

Communication (Rule 1.4)

Where AI use is material to a matter, clients may need to be told. The opinion frames this as part of keeping a client reasonably informed, not as a blanket disclosure on every use of every tool.

Supervision (Rules 5.1 and 5.3)

Partners and managers are responsible for firm-wide policy and for the work of the lawyers and nonlawyer staff using AI. Output is reviewed by a responsible lawyer before it leaves the firm, and that review is part of the supervisory system, not an afterthought.

Candor and meritorious claims (Rules 3.1, 3.3, 8.4)

AI does not lower the bar on candor to a tribunal or on filing non-frivolous claims. A hallucinated citation is the lawyer's problem, not the tool's.

Fees (Rule 1.5)

Fees stay reasonable. If AI saves time, you generally cannot bill for time you did not spend, and how you treat AI-related costs should be clear to the client.

Most firms do not fall short of Opinion 512 because they lack a policy document. They fall short because real AI use has quietly outrun the policy, the supervision, or the billing logic. Readiness is operational. A firm that can show what AI touched, who reviewed it, what the client was told, and how it was priced, on a real matter, is ready. A firm with a policy nobody follows is not.

That is why efficient.esq starts with an audit rather than a product. You map where AI is actually being used before you decide what to govern first.

Opinion 512 rewards firms that can produce a paper trail. If you can show the following, you have the core documentation in place.

• A current inventory of every AI tool in use, including general chat tools and AI features built into existing software, and where each one touches client matters.
• A risk classification for AI use cases (research, drafting, coding, client communication), with heavier controls on the higher-risk uses.
• A written AI policy that is actually maintained: approved tools, permitted tasks, prohibited data, and consent steps.
• The vendor confidentiality posture for any tool that sees client data, including whether inputs are retained or used for training, and a data processing agreement where appropriate.
• A supervision and verification record: who reviews AI-assisted work before it leaves the firm, and evidence that the review happened.
• Client communication and consent language, tailored to your approved workflows, for the situations where disclosure is warranted.
• A fee and billing approach that reflects AI time savings honestly.
• Training records showing lawyers and staff were trained before they were authorized to use AI on client work.
• An incident log for substantive AI errors or confidentiality concerns, and a record of what you did about them.

Before any policy, build the inventory. For each tool, capture what it is, who uses it, which tasks and matter types it touches, whether client data goes in, the vendor's data posture, and who reviews the output. The map is what turns an abstract duty into a list of specific decisions, and it is the input to everything else.