WASHINGTON, D.C.: Conversations with artificial intelligence tools could be used as evidence in court, U.S. lawyers are warning, following a key ruling that underscores the legal risks of treating chatbots as confidential advisers.

The warnings have intensified after a federal judge in New York ruled earlier this year that a former executive could not shield his chatbot interactions from prosecutors in a securities fraud case.

Legal experts say the decision highlights a critical distinction: while communications with lawyers are typically protected under attorney-client privilege, exchanges with AI platforms are not.

"We are telling our clients: You should proceed with caution here," said Alexandria Gutiérrez Swette, a lawyer at New York-based law firm Kobre & Kim.

Law firms across the United States have begun issuing advisories and updating client agreements to reflect the growing risks associated with AI use. More than a dozen major firms have outlined steps to reduce the likelihood that chatbot interactions could be disclosed in court.

Some have warned that sharing legal advice or confidential information with AI tools could undermine attorney-client privilege, which generally protects communications between lawyers and their clients.

The issue gained prominence in a case involving Bradley Heppner, former chair of bankrupt financial services firm GWG Holdings and founder of Beneficent. Heppner, who has pleaded not guilty to securities and wire fraud charges, used Anthropic's chatbot Claude to prepare reports related to his defense.

His lawyers argued that those materials should remain protected because they contained legal insights. Prosecutors countered that the chatbot interactions were not covered by privilege, since lawyers were not directly involved and AI tools do not qualify as legal advisers.

U.S. District Judge Jed Rakoff sided with prosecutors, ruling that Heppner must turn over 31 chatbot-generated documents.

No attorney-client relationship exists "or could exist, between an AI user and a platform such as Claude," Rakoff wrote.

The ruling marked an early test of how traditional legal protections apply in the era of generative AI. Courts are increasingly grappling with such questions as the use of AI expands in legal work and self-representation.

In a separate case the same day, U.S. Magistrate Judge Anthony Patti in Michigan took a different approach, allowing a woman representing herself to keep her ChatGPT conversations private by treating them as personal legal work.

ChatGPT and other generative AI programs "are tools, not persons," Patti wrote.

Despite that distinction, privacy concerns remain. Both OpenAI and Anthropic state in their policies that user data may be shared with third parties and advise users to consult qualified professionals before relying on chatbot guidance.

Rakoff also noted that Claude explicitly warns users they should not expect privacy in their inputs.

Law firms are now advising clients to limit what they share with AI systems, choose platforms carefully and, in some cases, explicitly state when chatbot use is directed by legal counsel.

Some firms suggest using language such as: "I am doing this research at the direction of counsel for X litigation," to strengthen claims of privilege.

Others are incorporating AI-related clauses into contracts. Sher Tremonte, for example, cautioned that "Disclosure of privileged communications to a third-party AI platform may constitute a waiver of the attorney-client privilege."

Lawyers expect further court decisions will clarify how AI-generated content is treated in legal proceedings.

Until then, many are urging clients to follow a familiar rule: avoid discussing sensitive legal matters with anyone outside their legal team, including AI tools.