A federal court in the Southern District of New York ruled that AI large language models such as Claude, ChatGPT, and Gemini are not attorneys, and communications with them are not privileged in criminal cases.

The decision signals a broad shift in how AI interactions are treated in legal contexts and highlights risks for individuals and businesses relying on AI for legal strategies.

Judge Jed S. Rakoff noted that Claude could only act as an attorney’s agent if an attorney directed the user to seek AI advice, which was not the case here.

The case centered on Bradley Heppner, former CEO of GWG Holdings, who discussed a defense strategy with Claude after becoming the target of federal investigations; prosecutors found 31 Claude conversations related to his case.

Industry reaction emphasizes that AI chats can be discoverable in court, and clients should avoid treating AI conversations as confidential or privileged.

The ruling states there is no trusting human relationship between a user and an AI, undermining attorney-client privilege for AI conversations.

The ruling rests on Claude not being an attorney and Claude’s terms of use allowing data retention that can be turned over to authorities in litigation.

During the proceedings, Claude stated it was not a lawyer and advised contacting a licensed attorney, yet the communications were deemed non-privileged.

Experts warn that this ruling de-emphasizes AI-assisted communications as confidential and may affect how executives and others use AI for legal planning.