In a recent US case, United States v. Heppner (No. 1:25 cr 00503 JSR), decided in the New York District Court, the court was asked to deal with a very modern legal question: what happens when someone involved in a criminal investigation communicates with a publicly available AI platform? More specifically, are those communications protected by the attorney-client privilege?

The short answer: no.

While the underlying facts of the case aren’t particularly important, what is relevant is what happened during the investigation. The Federal Bureau of Investigation (FBI) executed a search warrant at Mr Heppner’s home and seized various documents and electronic devices. Among the seized material were records of Mr Heppner’s interactions with an AI platform called Claude.

Mr Heppner’s attorney argued that these AI-generated documents should be protected by legal privilege. The reasoning was that the documents were created in anticipation of a possible indictment, and that Mr Heppner had input information into Claude that he had received from his lawyer. According to the defence, the purpose of using the AI platform was to help obtain legal advice.

In February 2026, the US Government asked the court to rule that these AI-related documents were not protected by the attorney-client privilege.

To put this in context, legal privilege generally applies to communications between a client and their attorney that are intended to be confidential and are made for the purpose of obtaining legal advice. In South Africa, there is an additional requirement: the legal advice must be sought in anticipation of litigation. The New York District Court found that the AI documents did not meet these requirements on several grounds.

• First, Claude is not a lawyer. As a result, communications with the AI could not be treated as communications between a client and an attorney.
• Second, Claude’s privacy policy explicitly states that the platform collects user data and may share it with third parties, including government authorities. Given this, the court found that Mr Heppner could not reasonably argue that his communications with the AI were intended to be confidential.
• Finally, the court held that Mr Heppner was not communicating with the AI platform for the purpose of obtaining legal advice in the sense required for privilege to apply.

In short, the court made it clear that using a public AI tool—even in connection with legal issues—does not attract the protection of the attorney-client privilege.

Ultimately, the Heppner decision is a useful warning for South African legal practitioners and clients alike: publicly available AI tools are no substitute for confidential legal advice. While AI platforms may assist with general research or preliminary thinking, they are not legal practitioners, and communications with them do not enjoy the protection of legal professional privilege. For individuals and businesses facing potential legal risk, the case highlights the need for caution when using AI and reinforces that sensitive information should be disclosed only to a qualified attorney, within a relationship that attracts privilege under South African law.