Are AI-Generated Materials Protected by the Work Product Doctrine?
Shealy v. Seaside Investments is one of the first Massachusetts decisions to address whether AI-generated materials are protected by the work product doctrine. Although its holding is narrow, it provides an early warning about the risks associated with clients’ independent use of generative AI.
In Shealy, upon receiving the defendants’ reservation of rights letter, the plaintiff emailed the letter and related documents to his girlfriend, who uploaded them to ChatGPT, a generative AI platform. Based on her prompt, ChatGPT generated a response letter and other materials, which she then emailed back to the plaintiff. No attorney participated in or directed this process. The defendants moved to compel production of the redacted email communications and attached documents, and the plaintiff asserted work product protection over those materials.
As the court explained, the work product doctrine protects (1) documents and tangible things, (2) prepared by or for a party or the party’s representative, (3) in anticipation of litigation or for trial. The court explained that the dispute centered on the second element: whether the AI-generated materials were prepared by or for the plaintiff’s representative.
The court first concluded that the plaintiff’s girlfriend was not the plaintiff’s representative for purposes of the work product doctrine. It then held that ChatGPT likewise did not qualify as the plaintiff’s representative. In reaching its conclusion, the court distinguished recent decisions out of the District of Colorado, Morgan v. V2X, Inc., and the Eastern District of Michigan, Warner v. Gilbarco. In both cases, the issue was the protection of materials prepared by pro se litigants. The court cited Morgan to explain that a pro se litigant is both the party and the advocate, and therefore, had better arguments for claiming work product protection in connection with their AI use.
Accordingly, the Shealy court concluded that neither the ChatGPT prompts nor the AI-generated output had been prepared by or for the plaintiff or a qualifying representative. Thus, because no attorney directed or participated in the AI use, the materials did not qualify for work product protection. The decision was also informed by the policy considerations underlying the work product doctrine. Under these circumstances, extending protection to the AI-generated materials would not advance the doctrine’s central purpose of safeguarding attorney preparation, mental impressions, and strategy.
Massachusetts is not alone in confronting this issue. In United States v. Heppner, a criminal case, the Southern District of New York held that materials generated through a defendant’s independent use of Anthropic’s Claude were not protected by the attorney client privilege or work product doctrine where the defendant acted on his own initiative.
These decisions serve as a reminder of the potential discovery risks associated with the independent use of generative AI in pending litigation. Litigants may assume that AI is simply another notetaking or brainstorming tool, but Shealy suggests that litigation-related AI use by parties may have consequences for whether resulting materials are discoverable. Although the decision leaves many questions unresolved, it counsels caution when litigants use AI for purposes related to pending litigation, particularly where that use occurs independently of counsel.
