In February 2026, the Honorable Jed S. Rakoff of the Southern District of New York decided yet another novel legal issue. This one relates to the use of AI, and may make you think twice before asking ChatGPT, Claude, or some other AI platform for legal advice. Judge Rakoff acknowledged that the case appeared to present a “question of first impression nationwide,” specifically: “when a user communicates with a publicly available AI platform in connection with a pending criminal investigation, are the AI user’s communications protected by attorney-client privilege or the work product doctrine?” The Court’s short answer is no – a finding that is sure to have wide-ranging and long-lasting implications throughout the modern legal landscape.
In late 2025, defendant Bradley Heppner received a grand jury subpoena that made it clear to him that he was the target of an investigation into securities and wire fraud. Without consulting with his counsel, Heppner turned to the generative AI platform, “Claude”, which he prompted to prepare reports outlining defense strategy and arguments based on the facts and his anticipated charges. Heppner was later indicted, and the FBI seized approximately thirty-one documents memorializing Heppner’s communications with Claude. The documents were seized pursuant to a valid search warrant. Heppner eventually sought to assert privilege over the seized communications with Claude, arguing that he “was preparing these reports in anticipation of a potential indictment,” inputted certain information learned from counsel, and shared the contents of the documents with counsel in furtherance of receiving legal advice.
The Court did not buy Heppner’s arguments. Beginning with the attorney-client privilege, Judge Rakoff noted that he could dispose of Heppner’s claim of privilege based solely on the fact that Claude is not an attorney. Further, Heppner could not have expected that his communications with Claude were confidential. This was not merely because Heppner communicated with a third-party platform, but because the written services policy, to which users of Claude consent to, provides that Anthropic (the program’s operator) collects data on both users’ inputs and Claude’s outputs, and Anthropic reserves the right to disclose the data to third parties. Moreover, the Court found that the disclosure policy put Claude users on notice that Anthropic may disclose user data “in connection with claims, disputes[,] or litigation,” even in the absence of a subpoena. Finally, in what the Court considered a “closer call”, it determined that Heppner did not communicate with Claude for the purpose of obtaining legal advice (within the meaning of the attorney-client privilege) because counsel did not direct him to do so and Claude explicitly informs users that it cannot give “formal legal advice”.
With respect to the work-product doctrine, which protects “materials prepared by or at the behest of counsel in anticipation of litigation or for trial”, the Court was similarly unconvinced. Given Heppner prepared these materials on his own, without direction from counsel, the Court found that he could not have been acting at the behest of his attorney. His counsel also conceded that they did not “reflect” the attorney’s strategy at the time of the communications. The Court then denied the remainder of Heppner’s arguments, which relied on distinguishable and non-binding authority.
As Judge Rakoff noted, “AI’s novelty does not mean that its use is not subject to longstanding legal principles, such as those governing the attorney-client privilege and the work product doctrine.” While this, and comparable issues, will surely continue to be litigated before various courts, it seems obvious that Judge Rakoff’s opinion will likely serve as a bedrock AI-related legal decision for years to come.
Bressler, Amery & Ross will continue to keep you apprised of this and other AI-related legal developments.