Can 'Non-Confidential' Discovery Materials be Uploaded? To an AI Tool for Analysis—Maybe, It Depends!

May 5, 2026

By: Mark A. Berman and Jessica L. Copeland

A recent decision by the U.S. District Court for the District of Kansas in Jeffries v. Harcros Chemicals held that litigants should not upload any discovery materials, whether confidential or nonconfidential, to open publicly accessible AI tools because such platforms may retain data, train on inputs and make deletion or clawback impossible. The court required the use of closed, secure AI tools with safeguards to prevent disclosure, preserve privilege and allow data destruction at the conclusion of the case, emphasizing risks to privacy, data‑protection compliance and discovery integrity. By extending these concerns to nonconfidential materials, the ruling reframes AI use as a structural discovery issue rather than a question of convenience and signals that efficiency gains do not outweigh the risks of losing control over litigation data.

Read more here.

This article was originally published on May 4, 2026, by the New York Law Journal.