Generative artificial intelligence tools have become increasingly prevalent across various domains of human activity. It has reliably been estimated, for instance that more than half of United States households have adapted AI in some form. Only three years after its release, one prominent AI platform is being used by more than 800 million people worldwide every week. Yet the implications of AI for the law are only beginning to be explored.
A ruling by Judge Jed S. Rakoff in United States v. Heppner (S.D.N.Y., Feb. 17, 2026) appears to be the first federal decision addressing whether a criminal defendant’s conversations with a generative AI platform are protected by attorney-client privilege or the work product doctrine. The answer on both counts according to this ruling was “no.” Most published decisions involving generative artificial intelligence have had to do with attorneys’ misuse of that technology. That set of concerns is plainly net present here.
Bradley Heppner was indicted on securities fraud and related charges stemming from an alleged $150+ million scheme involving GWG Holdings. After receiving a grand jury subpoena and learning he was a target, Heppner – on his own initiative, without his lawyer’s direction – used Claude to prepare roughly 31 documents outlining potential defense strategies and legal arguments. The FBI seized these “AI Documents” during a search of his home. Heppner’s counsel claimed privilege over them.
Heppner, through his counsel asserted privilege over these documents arguing that (1) Heppner had inputted into Claude, among other things,information that Heppner had learned from counsel; (2) Heppner had created the AI Documents for the purpose of speaking with counsel to obtain legal advice; and (3) Heppner had subsequently shared the contents of the AI Documents with counsel. Heppner’ s counsel conceded, however, that counsel “did net direct [Heppner] to run Claude searches.”
The trial court noted that it is well established that the attorney-client privilege attaches to, and protects from disclosure, “communications (1) between a client and his or her attorney (2) that are intended to be, and in fact were, kept confidential (3) for the purpose of obtaining or providing legal advice.” United States V. Mejia 655 F.3d 126, 132 (2d Cir. 2011). Courts construe the attorney-client privilege narrowly because it operates as an exception to the rule that “all relevant proof is essential” for a complete record and for “confidence in the fair administration of justice.”
On attorney-client privilege, the court found the documents failed on multiple independent grounds. First, Claude is not an attorney, so there was no attorney-client relationship. Second, the communications were not confidential — Anthropic’s privacy policy explicitly permits collecting user inputs and outputs, using them for training, and disclosing data to third parties including government authorities. Third, Heppner was not seeking legal advice from Claude; Claude itself disclaims the ability to give legal advice. The court noted that even though Heppner later shared the outputs with his lawyer, non-privileged communications don’t become privileged simply by being passed along to counsel.
On work product doctrine, the court held that even assuming the documents were prepared in anticipation of litigation, they were not prepared “by or at the behest of counsel.” Heppner acted entirely on his own. The documents didn’t reflect defense counsel’s strategy at the time they were created. The court distinguished a prior S.D.N.Y. magistrate decision (Shih v. Petal Card) that took a broader view, respectfully disagreeing and emphasizing that the doctrine’s core purpose is protecting lawyers’ mental processes, not a client’s independent research with an AI tool.
“Thus, the communications between Heppner and Claude were not privileged at the time they took place. Moreover, even assuming that Heppner intended to share these communications with his counsel and eventually did so, it is black-letter law that non-privileged communications are not somehow alchemically changed into privileged ones upon being shared with counsel. Thus, because the AI Documents would not be privileged if they remained in [Heppner’s] hands they did not acquire protection merely because they were transferred to counsel.” The court concluded that AI’s novelty doesn’t exempt it from longstanding legal principles.
The ruling has obvious implications for the millions of people using AI platforms to think through legal problems – those conversations are likely discoverable.