Court’s ruling offers cautionary tale for clients using generative AI

When a party communicates with a publicly available artificial intelligence (AI) platform in connection with a legal dispute, are the party’s communications protected by the attorney-client privilege or work product doctrine?

In one of the earliest decisions on this issue, Judge Jed S. Rakoff—an influential judge in the Southern District of New York—held that such communications were not protected. Although far from the final word on this issue, the decision offers practical lessons for clients seeking to preserve the confidentiality of their communications with their attorneys.

Facts

In 2025, Benjamin Heppner received a grand jury subpoena suggesting he had committed securities and other fraud. Shortly thereafter, he began using Claude, a generative AI platform, to prepare a defense strategy.

In late 2025, Heppner was arrested. In connection with the arrest, the FBI executed a search warrant at his home and seized, among other things, his electronic devices. His counsel later represented to the government that the seized materials included his communications with Claude regarding his defense strategy.

Heppner’s counsel asserted that the communications with Claude were protected by the attorney-client privilege and/or work product doctrine. The government disagreed and sought to review the communications.

Court’s decision