In February 2026, Judge Jed S. Rakoff held in United States v. Heppner that AI materials generated by a criminal defendant acting on his own volition were not protected. The defendant in Heppner had independently used a consumer AI tool without attorney direction. The Assini court distinguished those facts by emphasizing that attorney-directed AI usage—where counsel controls the prompts—triggers protection, while self-initiated use does not. The conflicting outcomes create immediate uncertainty: whether AI-generated materials are discoverable now depends on who initiated the interaction and whether an attorney directed the process.
Practitioners should treat attorney-directed AI work as opinion work product—the highest tier of protection—when counsel uses AI to research legal questions or draft arguments. The mental processes remain undiscoverable absent compelling need. However, materials generated by clients acting independently remain vulnerable. This creates a new discovery risk frontier. Attorneys must now distinguish between AI interactions they directed and those clients pursued on their own, and be prepared to argue jurisdiction-specific standards. The divergence between federal and state courts on this issue will likely persist until higher courts provide clearer guidance.