By: William D. Slicker
In a recent case of first impression, United States v. Heppner, Case No: 25-CR-503 (S.D.N.Y. 2026) the issue raised was whether the written exchanges between a defendant and artificial intelligence were protected from discovery by the plaintiff. The court ruled that they were not protected.
In October, 2025, a grand jury indicted defendant Heppner with securities fraud, wire fraud, and conspiracy to commit securities fraud and fire fraud, making false statements to auditors, and falsifying corporate records. The charges arose from Heppner’s conduct as executive of several corporate entities through which Heppner defrauded investors out of more than $150 million.
In November, 2025, F.B.I. agents executed a search warrant at Heppner’s home and seized documents and electronic devices. Among the seized materials were documents that were communications that Heppner had with AI platform Claude. Without any direction from his counsel, Heppner prepared reports that outlined defense strategy, that outlined what he might argue with respect to the facts and the law he anticipated that the government might use. Based on that, Heppner’s counsel argued that Heppner was preparing the reports in anticipation of litigation for the purpose of speaking with his counsel for advice, and therefore the communications were privileged under the attorney-client privilege or the work product privilege. The court rejected both arguments.
As to attorney-client privilege, the court stated that the attorney-client privilege protects communications (1) between a client and his attorney, (2) that are intended to be, and in fact are, kept confidential, (3) for the purpose of obtaining legal advice from his counsel. United States v. Mejia, 655 F. 3d 126 (2d Cir. 2011).
The court then applied those three elements to the Heppner circumstances. As to the first element, the court said the AI documents were not communications between Heppner and his attorney. Claude is not an attorney.
As to the second element, the court said that the written policy of Claude is that it uses the data to train Claude and it reserves the right to disclose the data to third parties. Claude also puts users on notice that even in the absence of a subpoena, it could disclose personal data to third parties in connection with claims, disputes, or litigation. For those reasons, Heppner had no reasonable expectation of confidentiality.
As to the third element, Heppner could not seek legal advice from Claude. When asked to give legal advice, Claude responded, “I’m not a lawyer and can’t provide formal legal advice or recommendations.”
Since the communications failed all three elements, they were not protected by the attorney-client privilege.
The court added that since the communications were not prepared by Heppner’s counsel or at his counsel’s request, the communications were not work product.
What does this ruling mean for your cases? If you or your client submit confidential communications, financial records, legal strategy, or anything else into ChatGPT, Claude, or any other large language model or AI platform, any privilege has been waived, and that information may be discovered by opposing counsel. In the future, you may expect AI communications to be raised in Requests for Production, Interrogatories, and Depositions.
Your initial conference with a client should now include instructions not to submit any communication between the attorney and client, any legal advice or strategy, any financial information, or anything else regarding the case to any AI platform. You should also instruct your staff to not input client data into AI platforms.
