The use of generative AI in the workplace is relatively new, but most professionals are already well aware of one of its major drawbacks: the information that AI chatbots provide is not reliably accurate. This means that it is extremely risky to use AI for legal purposes, because what the chatbot tells you the law is might very well be wrong.
But there is another significant red flag emerging in the use of AI in the legal industry – privacy. If you use ChatGPT, Claude, or any other public AI chatbot products to conduct your own research regarding a legal dispute, those AI conversations might later be discoverable in litigation. A federal court has issued one of the first rulings in the country on a question that every attorney and client using AI should be asking: is the information typed into an AI chatbot, and the output it generates, protected from disclosure? In United States v. Bradley Heppner, Judge Jed S. Rakoff for the U.S. District Court for the Southern District of New York answered no, at least on the particular facts before it.
The Heppner Decision Explained
Bradley Heppner was indicted in October 2025 on securities and wire fraud charges. He was aware that he was a target of a federal investigation and had already retained defense counsel. On his own initiative, without his lawyers’ direction, he used an AI chatbot to develop possible defenses. He later shared the results with his defense team. When Heppner was arrested in November 2025, the FBI seized devices containing those AI conversations. His defense identified 31 AI-generated documents and attempted to assert that they were protected by the attorney-client privilege and the attorney work product doctrine. On February 17, 2026, the court ruled that they were not protected.
The court gave three reasons why the AI-generated documents were not privileged. First, an AI tool is not a lawyer. The attorney-client privilege protects confidential communications with a licensed attorney for the purpose of obtaining legal advice. Communicating with a chatbot is not communicating with counsel, and the platform itself provided its information along with a disclaimer that it was not a lawyer and was not providing legal advice. Second, there was no reasonable expectation of confidentiality. The platform’s privacy policy disclosed that user inputs and outputs are collected, used to train the model, and may be shared with third parties, including the government. Third, the documents were not created to obtain legal advice. Heppner generated the documents himself to organize his own thinking, and the documents were not prepared as part of a communication to his lawyers.
The court also rejected the application of the attorney work product privilege, because Heppner prepared the documents on his own initiative rather than at counsel’s direction, and they did not reflect his lawyers’ mental impressions or strategy.
A Different Outcome: Warner v. Gilbarco
Just days before the Heppner decision, a federal magistrate judge in Michigan reached the opposite result on different facts. In Warner v. Gilbarco, Inc., a self-represented plaintiff used AI tools to help prepare her own litigation materials. The court refused to compel her to produce them, reasoning that AI programs are “tools, not persons,” so feeding material into one is not the same as disclosing it to an opponent. The court reasoned that because the work product privilege is waived only by disclosure to an adversary, her AI-assisted drafts were protected.
Key Takeaway
Heppner and Warner reached opposite outcomes, but there is a key distinction between the two cases in that the litigant in Warner was representing herself in the lawsuit, so she was able to avail herself of certain privileges that would not ordinarily apply to a layperson. It remains to be seen how New York courts will rule on this emerging issue. That being said, these two opinions still offer practical guidance that is clear enough to act on now: Resist the temptation to use an AI chatbot to brainstorm or work through the sensitive facts of your legal problem. Doing so risks that those AI conversations will eventually need to be disclosed to your adversary in litigation.
Legal Services
If you have questions about how your organization uses AI, or how to preserve privilege while doing so, contact us anytime.
(518) 855-3535
info@mclclaw.com
Citations: United States v. Bradley Heppner, No. 25-cr-00503-JSR (S.D.N.Y. Feb. 17, 2026); Warner v. Gilbarco, Inc. (E.D. Mich. Feb. 10, 2026).
———————————————————————————————————
This Client Advisory is provided as a courtesy to the clients of Monaco Cooper Lamme & Carr, PLLC. It provides general information and is not intended as legal advice and does not create an attorney-client relationship between MCLC and the reader. Should the reader desire additional information about the content of this Advisory and/or its application, please contact us at MCLC at (518) 855-3535 or info@mclclaw.com.
