Old Rules, New Tools: AI Prompts and the Limits of Privilege
Disclaimer: This post is provided for informational purposes only and does not constitute legal advice. Results depend on the specific facts and applicable law of each case.
Old Rules, New Tools: AI Prompts and the Limits of Privilege
Generative AI is now part of modern litigation. Lawyers use it to test claims and draft arguments. Clients sometimes use it before ever picking up the phone. The technology is new. The governing law is not.
Federal District Courts are starting to answer whether AI prompts and chat logs are discoverable, and all appear to point to the same conclusion, so far: Federal courts apply the pre-existing rules of privilege and work product.
In United States v. Heppner, 2026 U.S. Dist. LEXIS 32697 (S.D.N.Y. Feb. 17, 2026), a criminal defendant generated AI analyses on his own after learning he was under investigation. He later claimed the AI inputs and outputs were privileged. The court disagreed. The defendant failed to meet any of the necessary elements for privilege: the AI platform was not an attorney, the communications were not confidential (due to the platform’s terms of use), and the materials were not prepared at counsel’s direction. The opinion makes clear that independently generated AI content does not become privileged simply because it is later shared with a lawyer, but the ruling still allows for the possibility of privilege or work product to apply if an attorney is appropriately involved.
By contrast, in Tremblay v. OpenAI, Inc., 2024 WL 3748003 (N.D. Cal. Aug. 8, 2024), the court protected AI prompts crafted by counsel as opinion work product. Because the prompts reflected counsel’s mental impressions in preparing the case, they were “virtually undiscoverable.” The key distinction was attorney involvement.
Similarly, in Concord Music Group, Inc. v. Anthropic PBC, 2025 WL 1482734 (N.D. Cal. May 23, 2025), the court recognized that attorney-created prompts can qualify as work product; but warned that waiver principles still apply. A party cannot rely on selected AI outputs in its pleadings while shielding the broader testing process if fairness requires disclosure.
Finally, in Warner v. Gilbarco, Inc. (E.D. Mich. Feb. 2026), the court denied a broad request for AI-related materials, finding them not relevant or proportional. But that ruling turned on discovery standards, not the creation of a new AI privilege.
The common thread is simple: AI does not change the doctrine. Privilege requires communication with a real lawyer. Work product requires preparation by or at the direction of counsel. And waiver still occurs if a party discloses materials or places protected material directly at issue.
The tools may be evolving. The rules are not.
About the Author
Carl I. S. Mueller, Partner – Mueller Christensen LLP
Carl I. S. Mueller is a business litigation and professional liability defense attorney at Mueller Christensen LLP. He represents clients in complex civil litigation, malpractice disputes, fiduciary duty claims, and attorney fee disputes in Los Angeles, Orange County, and throughout California and Oregon, in both State and Federal courts. His practice focuses on high-stakes disputes involving businesses, attorneys, accountants, and fiduciaries.