By Jane Susskind
Lawyers now routinely use generative AI (“GenAI”). And because we’re lawyers bound by ethical rules, we approach GenAI use with healthy paranoia. We’re aware of the risks and take measures to protect our clients’ confidential information. Our clients are using GenAI, too, usually when litigation is nowhere on the horizon. Understandably, they might not be thinking about how their prompts and outputs could surface in discovery.
That leaves the discovery questions to us. Are prompts and outputs discoverable under standard discovery rules? What about other AI-generated materials, like notetaker transcripts and meeting summaries? And how should we counsel clients on the use and preservation of GenAI data? These questions are far from settled, but some courts have weighed in.
In re OpenAI, Inc., Copyright Infringement Litigation (“OpenAI Litigation”), a multidistrict litigation alleging that OpenAI used copyrighted materials without permission to train its models, is one of the leading cases on this topic. There, Magistrate Judge Wang ordered OpenAI to produce millions of ChatGPT logs, including prompts and outputs, and denied OpenAI’s motion to reconsider that order. OpenAI Litigation, No. 25-md-3143, 2025 WL 3468036 (S.D.N.Y. Dec. 2, 2025); see also OpenAI Litigation, 2026 WL 21676 (S.D.N.Y. Jan. 5, 2026) (overruling objection and affirming order compelling production). After analyzing the motion under traditional discovery principles, the court found the logs were relevant and proportional to the needs of the case, and any privacy concerns were mitigated by anonymizing the logs and the protective order. But in an earlier ruling in the same case, Magistrate Judge Wang denied a motion to compel the New York Times to produce logs reflecting its own GenAI use. OpenAI Litigation, 800 F. Supp. 3d 602 (S.D.N.Y. 2025). The New York Times argued that it would need to review over 80,000 log entries, which the court estimated would take over 1,300 hours and “cost upwards of $1 million for the privilege review alone.” Id. at 612. The court found that the logs were not relevant to the central issue in the case, but even if they were, production would not have been proportional to the needs of the case.
Takeaways
- Courts are treating GenAI like any other category of electronically stored information (“ESI”), so we should, too.
- Talk to clients about their GenAI use before a dispute, not after a discovery request.
- Explain to clients the importance of identifying which custodians use GenAI, which tools (including not only chatbots, but also notetakers and drafting assistants that sometimes run in the background), and where that data is stored.
- Amend your litigation hold letters to expressly address GenAI and preservation obligations related to chat histories.
- Broach the topic of GenAI data with opposing counsel early on and consider defining “Generative AI Data” in your ESI protocol.
While not an exhaustive list, these steps should put you in a better position when GenAI becomes an issue later in the case.
About the author
Jane Susskind is a member of McDonald Carano’s Commercial & Complex Litigation Practice and Appellate Practice. She is Co-Chair of McDonald Carano’s AI Task Force and has completed the “Generative AI for the Legal Profession” course provided by the Berkeley Law Executive Education program. Before joining McDonald Carano, Jane served for over two years as a judicial law clerk for Supreme Court of Nevada Justices Ron Parraguirre, Elissa Cadish, and Mark Gibbons.
About the article
This article was originally published in the Communiqué (Sep 2026), the official publication of the Clark County Bar Association.
The articles and advertisements appearing in Communiqué magazine do not necessarily reflect the opinion of the CCBA, the CCBA Publications Committee, the editorial board, or the other authors. All legal and other issues discussed are not for the purpose of answering specific legal questions. Attorneys and others are strongly advised to independently research all issues.
© 2026 Clark County Bar Association (CCBA). All rights reserved. No reproduction of any portion of this issue is allowed without written permission from the publisher. Editorial policy available upon request.