Skip to content
F·M FactMarshal
Menu
← Back to the docket
The FactMarshal Team ethics / compliance

ABA Opinion 512: six everyday practices it quietly made risky

On July 29, 2024, the American Bar Association issued Formal Opinion 512 — its first formal ethics guidance on lawyers’ use of generative AI. It didn’t ban anything. It did something more consequential: it mapped AI use onto six existing Model Rules, which means practices many firms treat as routine now carry compliance obligations most haven’t implemented.

The cautionary backdrop is real. In late 2023, a Colorado lawyer was suspended after filing a motion containing fabricated case citations generated by ChatGPT (People v. Crabill, Colo. O.P.D.J., Nov. 2023). The lesson of that case isn’t “don’t use AI” — it’s that the existing rules already reach AI-assisted work, and Opinion 512 spells out how.

The six pillars

1. Competence (Rule 1.1)

You can’t ethically use what you don’t understand. The opinion treats a reasonable understanding of the tools’ capabilities and limits — including their tendency to fabricate — as part of the duty of competence, and notes AI output “may be a starting point” but cannot be the end point. Practically: inventory the tools in use at your firm, train the people using them, and document both.

2. Confidentiality (Rule 1.6)

The opinion is blunt on one point: boilerplate technology-consent language in engagement letters is not adequate informed consent for feeding client information to generative AI tools. If client data touches an AI system, you need to understand where that data goes, and in many configurations you need informed consent that names the practice specifically. Vet tools before they touch client files; keep client data out of consumer tools entirely.

3. Communication (Rule 1.4)

Clients may need to be told when AI use is significant to the representation — and certainly when they ask. Establish disclosure triggers (AI materially shapes strategy or work product; the client inquires; billing reflects AI-assisted work) rather than deciding case-by-case under pressure.

4. Candor toward tribunals (Rule 3.3)

Every AI-generated citation and factual assertion gets independently verified before filing — no exceptions. A growing number of courts also impose their own AI-disclosure requirements by standing order, so track them per jurisdiction the same way you track local rules.

5. Fees (Rule 1.5)

If AI lets you do in two hours what used to take ten, you generally can’t bill the ghost eight. The opinion applies the familiar rule — fees must be reasonable and reflect actual time — to AI efficiency gains. This is the pillar with business-model consequences: it pushes AI-heavy work toward flat-fee and value-based arrangements, where efficiency is a margin advantage instead of a billing problem.

6. Supervision (Rules 5.1 / 5.3)

Managerial lawyers must ensure the firm has reasonable measures governing AI use — policies, an approved-tools list, training, and review procedures. Treat the tooling the way you’d treat a non-lawyer assistant: supervised, with a named owner.

States are not waiting

Opinion 512 is the baseline, not the ceiling. California issued practical guidance in November 2023; Florida’s Advisory Opinion 24-1 (January 2024) addresses consent for third-party AI tools; Pennsylvania’s joint formal opinion followed in 2024; Kentucky’s KBA E-457 (March 2024) takes a more permissive line on routine use. If you practice in multiple states, keep a jurisdiction matrix — the variance is real and growing.

A 90-day roadmap

  • Days 1–30: Inventory AI tools and users. Fix engagement-letter language. Issue interim use guidelines naming approved and prohibited tools.
  • Days 31–60: Adopt a written AI policy. Run training with sign-off. Stand up verification procedures for citations and factual assertions in filings.
  • Days 61–90: Review billing practices against Rule 1.5. Document everything — the firms that fare best under scrutiny are the ones that can show their process.

The deeper point

Opinion 512’s throughline is verifiability: understand the tool, verify the output, disclose honestly, bill honestly, supervise the process. That’s also a design standard for the tools themselves. AI whose every output is cited to a source page and passes through a documented human-review workflow doesn’t just make compliance easier — it makes the compliance story true. That’s the standard we hold FactMarshal to, and the one worth demanding from any tool that touches your cases.

Sources: ABA Formal Opinion 512, “Generative Artificial Intelligence Tools” (July 29, 2024); People v. Crabill, Colo. O.P.D.J. (Nov. 22, 2023); State Bar of California, Practical Guidance for the Use of Generative AI (Nov. 2023); Florida Bar Advisory Op. 24-1 (Jan. 2024); Pennsylvania Bar Ass’n Joint Formal Op. 2024-200; Kentucky Bar Ass’n Ethics Op. KBA E-457 (Mar. 2024). This article is commentary, not legal advice — read the opinions.