9
Every professional in a position of trust carries a duty that predates any technology: the obligation to protect the confidences of the people who rely on them. For attorneys it is the duty of confidentiality at the heart of the professional-responsibility rules. For accountants and tax practitioners it is the statutory and ethical obligation to keep client information confidential. For government contractors it is the handling of controlled unclassified information. The duty is not new. What is new is the ease with which a well-intentioned professional can now breach it without ever meaning to.
The duty does not pause for convenience
Artificial intelligence is a tool, and tools do not suspend obligations. The moment a professional pastes a client matter into a public AI service, they have transferred information that was entrusted to them into a system whose terms, retention, and access they do not control. The duty of confidentiality does not ask whether the transfer was convenient. It asks whether the professional took reasonable steps to protect the information. Convenience is not a defense; it is precisely the moment the duty is most at risk.
What the rules actually require
The professional-responsibility framework is built on competence and confidentiality. Competence now includes a working understanding of the technology a professional uses, including its risks. Confidentiality requires that a professional not reveal client information without informed consent, and that they use reasonable care to prevent its disclosure. When a firm adopts AI, both obligations extend to that adoption. The question is not whether the firm uses AI. The question is whether the firm can demonstrate that its use of AI protects the information it is duty-bound to protect.
Where the exposure hides
The risk is rarely the deliberate act. It is the ordinary one: a partner drafting a memo in a public tool, a paralegal running a conflict check through a free service, an associate summarizing a deposition with a consumer chatbot. Each is a single action that moves protected information outside the firm’s walls. The professional may believe the tool is private because it feels private. It is not. A public prompt is a transfer of the firm’s intellectual property and its clients’ confidences under terms the firm never read.
The standard is reasonableness, not perfection
No rule demands that a professional never use AI. The rules demand that a professional use it responsibly. That means knowing where the data goes, ensuring it is not used to train models the firm does not control, and being able to show that the firm’s AI runs in an environment the firm itself governs. When AI runs where the data already lives, the duty of confidentiality is not an obstacle to adoption. It is the reason the firm can adopt AI at all.
The firms that will lead are not the ones that avoid AI. They are the ones that use it in a way they can defend. The moat is not the model. It is the discipline of keeping client confidences inside the firm’s own walls, where the duty and the data both belong.
Start with the Confidential AI Exposure Checklist at 9tailsadvisory.com/checklist/. Nine lives. Zero leaks.