Court rules client’s pre-litigation AI chat logs were admissible as evidence of wrongdoing.
United States v. Heppner.
In October 2025, the United States charged Bradley Heppner with various counts of securities fraud, wire fraud, conspiracy, and falsifying corporate records. Heppner turned to AI for advice on what to do next, from outlining his overall defense strategy to how to present certain (bad) facts. On February 10, 2026, the Southern District of New York ruled Heppner’s chat logs were discoverable and could be used as evidence.
The Court’s Reasoning.
Heppner’s attorneys argued the chat logs were protected by the attorney client privilege and/or the work product doctrine. The court disagreed.
Attorney client privilege didn’t apply.
For a communication to be protected by the attorney client privilege under New York law (and under most state law), the communications must be (1) between the client and an attorney; (2) confidential; and (3) made to obtain or provide legal advice.
Here, the court held that Heppner’s communications with AI met none of these elements. First, AI isn’t an attorney. Second, the terms and conditions expressly stated that the information was not confidential and would be shared with third parties for, among other things, training purposes. And third, while Heppner claimed he was talking to AI to (arguably) obtain legal advice, the AI tool’s terms and conditions expressly stated that it (a) wasn’t a lawyer, and (b) couldn’t provide legal advice.
Work product doctrine didn’t apply.
For a document to qualify as “work product” under New York law, it must be prepared by or at the direction of counsel. Here, Heppner conceded that he voluntarily made the searches, not at the direction of counsel. The court indicated that the outcome may have been different had counsel directed Heppner to use the tool as part of litigation preparation.
Takeaway: Take Care when Consulting AI for Legal Advice.
Most states (including Texas) haven’t yet squarely addressed the discoverability of AI-generated materials in this context. In some jurisdictions, including Texas, material prepared in anticipation of litigation may qualify as work product regardless of whether directed by an attorney. You should consult with counsel regarding the applicable law in your jurisdiction.
AI is a powerful tool with many helpful uses. Here are our tips to protect your AI communications:
- Review AI platform privacy terms. Most consumer AI platforms automatically opt you in to sharing your searches for training purposes, but many allow you to turn that function off (potentially in exchange for a subscription fee). Disabling that function supports the argument that the communications were intended to be confidential.
- Engage counsel before you turn to AI. Under Heppner, your greatest exposure is before you engage counsel—because until then, your AI communications cannot have been made at counsel’s direction.
- Treat AI communications like emails or texts. Assume they may one day be read in court, because as Heppner shows us, they might be.
