Why this guide exists
Most lawyers are not afraid of AI. They are afraid of how to use it without violating their duties to clients. This guide names the duties, translates ABA Formal Opinion 512 and recent state-bar guidance into plain English, and gives you a checklist you can run against any AI vendor, including PracticeOps.
ABA Formal Opinion 512 in plain English
Opinion 512 (July 2024) addresses generative AI under Model Rules 1.1 (competence), 1.6 (confidentiality), 1.4 (communication), 5.1/5.3 (supervision), 1.5 (fees), and 3.3 (candor). Below is each duty in one sentence, with the practical implication for a solo or small firm.
- Competence (1.1): You must understand the AI well enough to evaluate its outputs.
- Confidentiality (1.6): Do not put client information into a system that uses it for training or shares it without consent.
- Communication (1.4): Tell clients when AI materially shapes their representation.
- Supervision (5.1/5.3): AI is treated like a non-lawyer assistant; you supervise it.
- Fees (1.5): Bill based on the value to the client, not on hours saved by the AI.
- Candor (3.3): Verify citations before filing.
State-bar variations to know
(Fabio to expand: New York, California, Texas, and Florida, the top 4 markets for the initial cohort. Cite the controlling opinion and date for each.)
Five questions to ask any AI vendor
- Where does my client data live, and who else can read it?
- Is my data used to train your models or anyone else's?
- What audit trail will I have if a bar grievance arrives in two years?
- Who is accountable when the output is wrong?
- Is your pricing tied to my hours, my fees, or a flat retainer?
Privacy architectures that satisfy these duties
Several architectural patterns can satisfy the duties above. PracticeOps uses one of them: your agent runs on a private server your firm owns, your data never leaves it, and it is never pooled with other firms or used to train any model. Other valid patterns include on-premise deployments, single-tenant cloud instances with contractual no-training clauses, and BYO-key arrangements with major model providers. The pattern is less important than the answers to the five questions above.
Sources and further reading
- ABA Formal Opinion 512 (July 29, 2024)
- State Bar of California, Practical Guidance for the Use of Generative AI (Nov 2023)
- Florida Bar Ethics Opinion 24-1 (Jan 2024)
- New York State Bar Task Force on AI Report (Apr 2024)