Use AI on client matters without revealing client confidences.

Model Rule 1.6 protects information relating to the representation, and ABA Formal Opinion 512 (2024) applies it to generative AI: lawyers must evaluate the confidentiality risk before putting client information into these tools, and need the client's informed consent before using a self-learning tool with it — boilerplate consent in an engagement letter is not enough. The Pennsylvania Bar Association and Philadelphia Bar Association's Joint Formal Opinion 2024-200 reaches the same conclusions and adds practical guidance.

Saibre changes what the AI receives rather than asking you to trust what happens afterwards: client names, matter numbers, and account details become placeholders on your machine. The local protection log gives you something concrete for the file. You remain responsible for your professional obligations — this supports them; it does not discharge them.

A confidentiality lapse is a Rule 1.6 matter for your disciplinary board, a potential malpractice claim, and, because disclosure to a third party can put privilege at risk, a problem inside the case itself. None of those comes with a dollar cap, and none of them requires a breach: the disclosure is the act.

Opinion 512 puts the burden on you to know how a tool handles data before client information goes in. With Saibre the answer is short: the identifiers never went in.