A federal judge has ruled that transcripts of conversations between a criminal defendant and the artificial intelligence chatbot Claude are not protected by attorney-client privilege, a decision that establishes a significant legal precedent for the use of AI tools in legal matters. The ruling, delivered Tuesday by Judge Jed Rakoff of the Southern District of Novel York, came during a pre-trial conference in the case of Bradley Heppner, a Dallas financial services executive accused of securities fraud and wire fraud.
Heppner allegedly used Claude, developed by Anthropic, to research legal questions related to the government’s investigation following the receipt of a grand jury subpoena and before his arrest on November 4, 2025. Federal agents seized approximately thirty-one documents generated by the AI during a search of Heppner’s mansion, prompting a dispute over their admissibility as evidence. Defense counsel argued the materials were privileged communications, but prosecutors countered that the AI-generated content lacked the protections afforded to traditional attorney-client interactions.
The government’s motion rested on three primary arguments. First, prosecutors asserted that because Claude is not an attorney and no attorney participated in the creation of the documents, the communications could not qualify for privilege. Second, they pointed to Anthropic’s terms of service, which explicitly disclaim the provision of legal advice. Finally, the government argued that Heppner’s voluntary submission of queries to a third-party platform with a privacy policy allowing data collection and potential disclosure to authorities waived any claim of confidentiality. According to the government’s filing, “The defendant appears to have directed legal and factual prompts at an AI tool, not his attorneys.”
Judge Rakoff agreed with the government’s assessment, finding no basis for asserting either attorney-client privilege or work product protection for the AI-generated documents. The government further argued that even if the documents were subsequently transmitted to counsel, this act did not retroactively establish privilege. The judge indicated a formal written order detailing the ruling would follow.
Legal experts note that the decision relies on established principles of law, rather than creating entirely new legal doctrines. However, the case highlights the novel challenges posed by the increasing use of AI in legal practice. Andrew R. Lee, Jason M. Loring, and Graham H. Ryan, attorneys at Jones Walker LLP, wrote in a client alert that the ruling confirms what many suspected: “conversations with AI tools are not protected by attorney-client privilege.”
The ruling does not preclude the possibility of AI-generated documents being protected in all circumstances. The applicability of privilege and work product protections will likely be determined on a case-by-case basis, considering the specific facts and jurisdiction. Attorneys advising clients on the use of AI tools are cautioned to understand the policies and parameters of those tools, and to carefully document their use, including the purpose of queries and the involvement of counsel.
The issue of AI-generated evidence and privilege was identified as one of nine key challenges in a recent cross-border white collar crime and investigations review conducted by A&O Shearman, underscoring the growing importance of this legal area. The firm’s disputes practice is closely monitoring global developments on the issue, and is available to advise businesses on navigating the evolving legal landscape.
