How many times have you asked ChatGPT or Google Gemini to draft an email, outline a motion, or summarize a case? In 2025, generative AI tools are increasingly commonplace in law firms, corporate legal departments, and companies across virtually all sectors. They can promote efficiency, lead to cost savings, and optimize workflow. Such tools, however, should be used responsibly, especially when litigation looms. There is seemingly one question no one can ignore: How do we grapple with confidentiality and privilege issues surrounding AI chat histories and their admissibility in court?
The answer may be far from settled, but the trajectory is clear. AI chat histories will be introduced as evidence in litigation, but their admissibility and legal standing are still being defined. See Laurence Hammack, Roanoke Teen Convicted of Murder Consulted AI Chatbot, Roanoke Times (Oct. 20, 2024). The integration of AI into daily life means that these digital chat logs, which feel private, can become discoverable records subject to legal scrutiny. Courts continue to grapple with how to treat AI chat histories, creating both challenges and opportunities for the legal community. See, e.g., Matter of Weber, 85 Misc. 3d 727, 742 (N.Y. Sur. Ct., 2024) (“The mere fact that artificial intelligence has played a role, which continues to expand in our everyday lives, does not make the results generated by artificial intelligence admissible in Court”).
Use in Litigation: It’s All Data
On closer examination, courts are likely to treat AI interactions as electronically stored information (ESI), akin to emails or text and WhatsApp messages. If so, our carefully crafted AI prompts and the bot’s responses could become the next “Exhibit A” in a courtroom. Under Federal Rule of Civil Procedure 34(a), parties may request discovery of any document, “including writings, drawings, graphs, charts, photographs, phonorecords, and other data compilations.” The inclusive description of the term “document” accords with changing technology. Courts therefore interpret ESI broadly to include, for example, emails, instant messages, platform logs, metadata, and digital records. See Zubulake v. UBS Warburg, 217 F.R.D. 309, 316-17 (S.D.N.Y. 2003). AI chat histories arguably fit within such a definition, making them similar to business chat logs. If those logs contain relevant information, expect opposing counsel to come knocking.
A separate but equally as important layer to this concerns the discoverability of such records. Rule 26 provides, in part, that discovery must be “relevant to any party’s claim or defense” and “proportional to the needs of the case.” Fed. R. Civ. P. 26(b)(1). AI chat histories can provide insight into a user’s intent, knowledge, and thought processes at a specific moment in time. In corporate litigation, they can potentially reveal business strategies, internal deliberations, or discussions about a client. On the employment front, consider the implications of AI chat queries such as “Can we fire someone for XYZ?” or “Is it legal to terminate an employee for XYZ?” If a compliance officer asked an AI chatbot, “Can I legally do this under SEC rules?” such an exchange might be highly relevant under Rule 26.
Now comes the elephant in the room. Whether and to what extent AI chat histories can be withheld or precluded on the basis of an applicable evidentiary privilege is yet to be determined. In light of this uncertainty, the most prudent course of action would be to avoid interactions with AI tools that involve privileged communications or information. Absent that, users should bear in mind some considerations on the topic. The client wields the power to either assert or waive the attorney-client privilege. See In re Lindsey, 158 F.3d 1263 (D.C. Cir. 1998). When a lawyer voluntarily shares confidential client information with an AI chatbot, there is arguably a waiver of privilege. The outcome is likely the same if “a client inputs data into ChatGPT [or other AI tool] while represented by an attorney.” Jordyn C. Tye, Comment, Exploring the Intersections of Privacy and Generative AI: A Dive into Attorney-Client Privilege and ChatGPT, 64 Jurimetrics J. 309, at 332 (2024). As previously identified, however, the issue of privilege and our interactions with AI tools remains unsettled.
The Duty to Preserve Evidence Includes AI Chats
A duty to preserve evidence is triggered when there is knowledge of a potential claim. Winters v. Textron, 187 F.R.D. 518 (M.D. Pa. 1999). A potential claim will be considered cognizable when a party reasonably believes that litigation is foreseeable. Once such a duty arises, a party must promptly enforce a litigation hold to preserve potentially relevant evidence. See Crown Castle USA v. Fred A. Nudd, 2010 WL 1286366, at *10 (W.D.N.Y. Mar. 31, 2010) (observing that the “failure to implement a litigation hold at the outset of litigation amounts to gross negligence”). The litigation hold will necessarily include AI chat logs and histories. For this reason alone, internal preservation and destruction policies should be updated to reflect AI chat data. The deletion of such records can leave the door open for sanctions due to spoliation of evidence. This could include an award of attorneys fees, the outright dismissal of a claim, or an adverse inference instruction.
By extension, adopting sensible AI governance policies can keep you ahead of the curve as the technological landscape continues to evolve. Attorneys, staff, and team members should be apprised of acceptable uses of AI tools, including chatbots, corresponding security requirements, as well as retention rules. Words of caution should also be shared about the use of sensitive or confidential client information with AI tools.
The Bottom Line
AI and AI tools are here to stay. From a macro perspective, we should all be mindful of our interactions with such tools and appreciate the possibility that chat histories are likely to become a regular part of discovery. The challenge for lawyers is not just a technical one; we should strive to integrate AI tools in a strategic and ethical manner. Lapses in judgment or a lack of training and education on the topic can lead to your late-night ChatGPT query being introduced by the plaintiff as “Exhibit B.”
Reprinted with permission from the September 12, 2025 edition of The New Jersey Law Journal © 2025 ALM Global Properties, LLC. All rights reserved. Further duplication without permission is prohibited, contact 877-256-2472 or reprints@alm.com.