On April 16, Reuters reported that as people increasingly turn to AI for advice, some lawyers in the United States are warning their clients not to treat AI chatbots as trustworthy confidants when it comes to personal freedom or legal liability.


Don’t tell the AI ​​everything

Now those warnings are taking on added urgency after a federal judge in New York ruled this year that the former CEO of a bankrupt financial services company could not prevent prosecutors from accessing his AI chat logs. Prosecutors have filed securities fraud charges against him.

In the wake of the ruling, lawyers have been reminding clients:In criminal cases, prosecutors may request records of conversations between the client and chatbots such as Anthropic’s Claude and OpenAI’s ChatGPT. In civil cases, opponents may also make similar claims.

“We tell our clients: You should be careful about how you chat with AI,” said Alexandria Gutiérrez Swette, a lawyer at the New York law firm Kobre & Kim.

Under U.S. law, discussions between people and their attorneys are almost always considered confidential. But AI chatbots are not lawyers, so lawyers are instructing clients to take steps to keep their communications with AI tools more private.

In emails to clients and announcements posted on their websites, more than a dozen major U.S. law firms have outlined recommendations for individuals and companies to reduce the likelihood that AI chat logs will end up in court.

Similar warnings have begun to appear in cooperation agreements signed between some law firms and clients. For example, New York-based Sher Tremonte made it clear in its recent client contracts that inputting a lawyer's advice or communications into a chatbot could invalidate the originally protected attorney-client privilege, which is typically used to protect attorney-client communications from being disclosed.

warning jurisprudence

The case that raised the alarm involves Bradley Heppner, the former chairman of bankrupt financial services company GWG Holdings and the founder of alternative assets firm Beneficent. Heppner was indicted by federal prosecutors in November on securities fraud and wire fraud charges and has pleaded not guilty.

Hepner had used Anthropic's chatbot Claude to prepare reports on his case to share with his attorneys. Lawyers later argued that his communications with the AI ​​should not be made public because they contained information from lawyers defending the case.

But prosecutors argued they were entitled to request access to the material Hepner generated using Claude because the lawyer was not directly involved in generating the content and the attorney-client privilege did not apply to chatbots.

For a client, voluntarily disclosing information provided by an attorney to any third party may jeopardize the legal protections normally afforded to those attorney communications.

In February, U.S. District Judge Jed Rakoff in Manhattan ruled that Hepner must turn over 31 documents related to the case that were generated by Anthropic's Claude.

Rakoff believes that there is no and cannot be an attorney-client relationship between AI users and platforms like Claude.

As of press time, Heppner's attorney had not commented. A spokesman for the U.S. Attorney's Office in Manhattan declined to comment.