Presented with a question of first impression, the United States District Court for the Southern District of New York in United States v. Heppner, No. 25-cr-00503-JSR (S.D.N.Y. Feb. 17, 2026), held that a defendant’s use of AI in connection with a pending criminal investigation was not protected by the attorney-client privilege or work product doctrine.

A grand jury indicted Bradley Heppner, an executive of several corporate entities including a publicly traded company, with securities fraud, wire fraud, conspiracy to commit securities fraud and wire fraud, making false statements to auditors, and falsifying corporate records. In connection with his arrest, the FBI seized documents and electronic devices, including thirty-one documents memorializing communications that Heppner had with Claude, a generative AI platform. Without any request or suggestion from his attorney, Heppner prepared reports outlining defense strategy, including what arguments he could make in response to facts and law the government might be charging. Heppner used the AI platform after he received a grand jury subpoena and discussions with the government showed he was the target of its investigation.

In exchanges with the government, Heppner’s counsel asserted privilege over the AI documents. His attorney argued the documents were protected because Heppner had inputted information he learned from counsel into Claude, created AI documents to obtain legal advice, and shared them with counsel. However, the Court held that two (if not all three) elements of the attorney-client privilege were not present. Perhaps most importantly, Heppner’s claim of privilege was swiftly denied because Claude is not an attorney. Further, because Claude’s privacy policy permits the use of information entered in the platform to train the bot and allows disclosure to third parties, there could be no expectation of confidentiality. Finally, Heppner did not communicate with Claude to obtain legal advice. Although he used Claude for the purpose of talking to counsel (such that Claude could have arguably functioned as a highly-trained professional acting as the lawyer’s agent), he did not do so at the request or direction of counsel, and providing the documents to counsel did not retroactively protect their contents.

Heppner also sought to protect the AI documents under the work product doctrine. However, the Court quickly disposed of this argument because, even assuming they were prepared in anticipation of litigation, Heppner’s counsel admitted they were not prepared by counsel nor did he request his client seek such information from Claude. Rather, Heppner prepared the AI documents on his own volition. Further, counsel also conceded that although the AI documents affected his strategy, they did not reflect his defense strategy when they were created.

The Court concluded its opinion by stating that “…Al’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 the Court acknowledged that “[g]enerative artificial intelligence presents a new frontier in the ongoing dialogue between technology and the law,” the AI documents generated from Heppner’s use of Claude did not warrant protection under the attorney-client privilege or work product doctrine.

Although the Court in United States v. Heppner did not specifically address whether AI platforms with stronger data protection and confidentiality safeguards could be treated differently, the decision serves as a strong warning that courts faced with similar facts will likely analyze the status of AI-generated documents under the well-settled principles on which the attorney-client privilege and work product doctrine are based.

What are the practical ramifications of this opinion? Documents generated with an AI platform will not be protected under the attorney-client privilege by way of a client simply providing them to their attorney. Further, clients should be instructed not to input any information learned from their attorney about their case into an AI tool as that destroys confidentiality. Finally, protection under the work product doctrine requires attorney direction – but proceed with caution. As the law on the use of AI develops, clients and lawyers alike cannot expect that documents generated by AI, regardless of who prepares them, will be protected from discovery at this time. Law firms should adopt strong internal AI policies to address the inherent risks of artificial intelligence and advise their clients concerning AI use through robust engagement letters to avoid waiving privilege.