Federal Court Rules AI Interactions Are Discoverable
On February 17, 2026, a Manhattan federal judge in United States v. Heppner answered a question of first impression: whether AI interactions are discoverable in litigation. The court ruled the defendant’s AI interactions that he prepared and shared with his attorneys are subject to discovery. The court further concluded that he cannot claim attorney-client privilege or work product doctrine to avoid producing them to the other side.
Given AI’s widespread use today, the implications reach into nearly every workflow and business operation. Individuals and businesses should exercise caution using AI chatbots like ChatGPT and Claude when handling work or dispute strategy, as those communications and AI-generated materials may ultimately have to be disclosed in court.
Case Summary
In October 2025, the federal government indicted Heppner in connection with an alleged $150 million investor fraud scheme. After receiving a grand jury subpoena and learning he was a target of an investigation, Heppner prepared reports using AI to outline his defense strategy and arguments relating to facts and law he anticipated he would be charged with. Importantly, Heppner did this on his own volition, not at his attorneys’ request.
When the FBI later arrested Heppner and seized these documents, Heppner’s counsel asserted privilege over the AI documents. Counsel argued that he input information learned from them, created the documents to obtain legal advice, and subsequently shared them with counsel. The government moved to compel production of the AI documents, asserting they are not protected. Surprisingly to many nonlawyer AI users, the court agreed.
Attorney-client privilege protects communications with an attorney made to seek legal advice. AI chatbots are not attorneys. Therefore, communications with them are not kept confidential, particularly where the platform’s policy provides that inputs and outputs are used to train the platform which reserves the right to disclose them to third parties. The communications with AI tools are treated the same as any other media such as internet searches, emails, and text messages. Sharing counsel-derived information with AI may constitute disclosure to a third party and waive privilege.
The court also rejected the attorney work product doctrine. That doctrine generally protects an attorney’s mental processes to analyze and prepare a case. The attorney work product doctrine can apply to nonlawyer client created materials, but only if “prepared by or at the behest of counsel in anticipation of litigation or for trial.” Defense counsel conceded that Heppner created the AI documents independently, and they did not reflect his attorneys’ strategy when created.
Key Takeaways
- Anything entered into or received from AI tools may be discoverable and is unlikely to be privileged.
- Request AI materials in discovery, and if necessary, subpoena AI companies for them where relevant.
- Document attorney direction if an attorney wants a client to create documents (AI-generated or not). While this does not ensure protection against producing documents, it could demonstrate that the work was performed at counsel’s request.
- Engage counsel directly when developing legal strategy. AI platforms do not provide the protections afforded by attorney-client privilege.
Contact
If you have questions about the implications of this ruling, please contact Dan Lampke or your Bernkopf attorney.
