What Opinion 512 actually requires.

In July 2024 the American Bar Association issued its first ethics opinion on generative AI. Formal Opinion 512 is usually summarized as a warning. Read closely, it is something more useful: a short list of specific duties, most of which a firm can satisfy with ordinary discipline.

The opinion applies existing rules rather than inventing new ones. Competence means understanding, at a practical level, what a tool does with what it is given. The duty of confidentiality under Rule 1.6 applies in full to anything entered into an AI tool. And for tools that learn from what users type — the opinion calls them self-learning — a client’s informed consent is required before information relating to the representation goes in. Boilerplate consent in an engagement letter, the opinion says, is not sufficient.

Two things Opinion 512 does not do. It does not prohibit cloud tools; providers can be used under the same reasonable-efforts vendor diligence that has governed cloud computing for a decade. And it does not require any particular architecture.

What it does is assign homework. A firm needs to know which tools are in use, what each one does with its inputs, which matters may touch them, and when consent is required. Where no client information enters a third-party tool at all, the consent question never arises — one reason some firms are moving this work into systems of their own.