New York bill would protect confidential legal work handled by AI

Confidential legal communications and attorney work prepared with AI would keep their protected status. This applies when AI handles the material, including AI run by a third party or on a private server. The communication or work must be confidential and made to get legal advice or strategy, or to prepare or help prepare legal materials. The bill would take effect immediately if enacted.

Before using AI for a client’s legal work, an attorney would have to provide a signed and notarized consent form. The form is required for attorneys using AI for the legal purposes described in the bill. The form must explain risks such as made-up or inaccurate information, inadequate review, and poor legal strategy. The form would apply from the date the client first engages the attorney, regardless of when it is signed or notarized.

An attorney’s AI use for a client would require human review; a form allowing use without review would be void, and that client’s AI-handled information would lose its protected status and become obtainable. The human-review rule and loss of protection apply to AI use for a particular client. A client would not be responsible for an attorney’s AI errors, except when the attorney is representing themself. Consent would not prevent a client from bringing a malpractice claim, challenging ineffective assistance, filing a disciplinary complaint, or raising a related matter. People representing themselves would not need a form, and the protections would presumptively apply to them. The bill would take effect immediately if enacted.