In United States v. Heppner, Judge Rakoff of the Southern District of New York held that documents outlining legal strategy that a defendant created using Anthropic’s Claude and shared with counsel were not protected by the attorney-client privilege or the work product doctrine. While this decision has received widespread and at times critical attention, it applied traditional principles: Claude is not a lawyer, the client did not consult Claude at the direction of counsel, and Anthropic’s privacy policy gave the client no reasonable expectation of confidentiality.

Whether Heppner correctly decided this issue or not, there are many other issues posed by AI that will become increasingly important to lawyers and their clients as AI pervades the practice of law and corporate life. Here, I highlight one such issue that could prove particularly important in future litigation and investigations: the privilege risk of lawyers using AI notetaking tools such as Otter.ai and Zoom AI Companion to create recordings, transcripts and summaries of their meetings. Unlike Claude as used in Heppner, notetaking tools do not generate original content, but instead raise the potentially even more serious concern that sensitive attorney conversations, once memorialized, may be divulged.

When AI notetakers are used in attorney-client meetings, one risk is inadvertent waiver. The attorney-client privilege protects only confidential communications, and introducing a tool that stores content on third-party servers and uses it to train machine learning models may threaten confidentiality, as the New York City Bar Association recognized in its December 2025 ethics opinion on AI notetaking tools. The dangers resulting from waiver of a verbatim transcript or even a summary are significant, as these documents could reflect sensitive discussions without being filtered through a lawyer’s judgment as to what is important and appropriate to memorialize.

Even where the privilege remains intact, an adversary may argue that it does not extend to the entirety of the AI transcript or summary. Lawyers routinely wear multiple hats, including advising on business strategy, and statements reflecting business advice are generally not considered privileged. Once AI-generated notes are created, they will be subject to review in litigation, and non-privileged portions will have to be produced in discovery. Battles over redaction of such documents will arise and pose risks and costs to parties and their counsel alike.

The risks are different but still significant when AI notetakers are used in conversations that are not themselves privileged, such as interviews of third-party witnesses. Lawyers can often shield their own notes of such conversations under the work product doctrine, which protects materials reflecting counsel’s mental impressions and strategies. But an AI-prepared transcript or summary may prove more difficult to protect as attorney work product.

Companies and law firms should consider what practices they wish to follow to address the privilege and work product risks posed by AI notetakers. It is critical to adopt policies in this complex and evolving landscape before documents are created that could become the subject of litigation years later.

--------

This publication is for general information only, does not create an attorney-client relationship and should not be relied upon as legal advice. © 2026 Gruenstein Law PLLC.