Supervised AI Use

ABA Model Rules 5.1 and 5.3 were drafted for a pre-AI bar. Both rules survived intact into the AI era because the ABA Standing Committee on Ethics and Professional Responsibility, in Formal Opinion 512, treated AI tools as a non-lawyer “assistant” subject to Rule 5.3 supervision and treated the firm’s adoption of those tools as a managerial-supervision question under Rule 5.1. That dual reading is now the structural basis for the state-bar opinions issued to date on attorney AI use and for most carrier-side renewal questionnaires.

The supervisory duty has three operational pieces. First, the firm needs a written policy that names the approved tools, prohibits the unapproved ones, and identifies who supervises usage. Second, every attorney and staff member using an approved tool needs documented training, refreshed when the tool changes in a material way. Third, the firm needs a register that records who used which tool on which matter, so a partner can answer the supervisory question after the fact. None of this is a new ethics theory. It is what Rule 5.3 has required for paralegal supervision since the 1980s, applied to a tool that talks back.

State bars have been broadly aligned on the supervision framing, with state-specific overlays. California’s CGAO Practical Guidance, Florida Ethics Opinion 24-1, and the New York State Bar’s 2024 report all adopt the supervision-of-non-lawyer framing for generative AI. Where they differ is on what documentation the supervising lawyer needs to produce in a discipline proceeding. The state tracker captures the specific rule-by-rule overlays; the policy template and training resources here produce the documentation those rules expect.

Carriers have moved in parallel. The seven-question underwriter checklist in the AI liability insurance guide is, in substance, a Rule 5.1 audit. A firm with a documented program already passes it.