CLIENT ALERT: Think Your AI Chats Are Private? They May Be Discoverable in Court

As AI continues to gain steam in all aspects of life, clients increasingly turn to AI programs like ChatGPT, Claude, Gemini, or Co-Pilot before consulting their attorneys. Amidst this trend, a federal court in New York has issued on Feb. 17, 2026, what appears to be the first judicial ruling directly addressing whether client-initiated, AI-generated content is protected by attorney-client privilege or work product doctrine.0F0F[1] The answer is no. The decision has immediate implications for anyone who uses publicly available AI tools in connection with actual or anticipated litigation, regulatory matters, or sensitive business disputes.  While consulting AI may seem like a good idea at first, whether to calm concerns or as an effort to save time, anything a client puts into these AI programs may be discoverable. Conversations with AI chatbots are not confidential in the way that communications with your lawyer are.

The court's analysis turned on five straightforward points that apply broadly to publicly available AI platforms, including:

1. AI is not an attorney, and no attorney-client relationship can exist between a user and a platform like Claude or ChatGPT.
2. Users have no reasonable expectation of confidentiality in their AI conversations because the program's own privacy policies expressly permit the collection and disclosure of user inputs and outputs.
3. Even where a user inputs information received from counsel into an AI platform, doing so constitutes a voluntary disclosure to a third party that waives privilege over that information, just as sharing privileged communications with any outside party would.
4. Changing the privacy settings or using a paid AI service does not necessarily render the information protected from discovery. The analysis may be different if an attorney uses the AI platform, or directs the client to do so, and the user selects an enterprise AI platform whose terms of service maintain confidentiality of user inputs. Outside that narrow context, assume that anything input into an AI tool may be discoverable or disclosed.
5. AI-generated documents created by a client on their own initiative and without explicit direction from counsel are not privileged, even if the client later shares those documents with their attorney.

This means that if you use an AI tool to draft strategy memos, analyze legal exposure, or organize facts related to a dispute, those materials may be fully discoverable. More critically, if you input a law firm's draft documents, legal memoranda, or privileged communications into a public AI platform, you risk waiving the privilege and protection that would otherwise attach to those materials if you had kept this flow between you and an attorney.

Many clients turn to AI for legal strategy before even consulting an attorney, a pattern that is becoming increasingly common, but which this new decision makes considerably more risky.  While the decision leaves open whether enterprise-grade AI platforms, or consumer platforms used at the direction of counsel, might afford greater protection, those questions remain unsettled. The prudent course to mitigate these risks is to consult your attorneys before consulting an AI platform for any matter involving actual or anticipated litigation, regulatory exposure, or sensitive business information. The cost of a brief conversation with your attorney is far lower than the cost of producing your own strategy or confidential business information to opposing counsel.

[1] United States v. Heppner, ---F.Supp.3d---, 2026 WL 436479, at *1 (S.D.N.Y. Feb. 17, 2026).