What US v. Heppner actually held.

In February 2026, Judge Jed Rakoff of the Southern District of New York ruled on a question many firms had treated as theoretical: whether a person’s conversations with an AI chatbot are protected the way conversations with a lawyer are. They are not.

The defendant had used a consumer chatbot to think through matters that later became the subject of a federal prosecution. The FBI seized roughly thirty-one of those conversations. He asserted attorney-client privilege and work-product protection. The court held that neither applied: the conversations were not with counsel, were not made at counsel’s direction, and took place in a consumer service, outside any relationship the law protects. The seized chats could be used against him.

The holding is narrower than the headlines. Heppner does not say that using AI destroys privilege; it says privilege never attached in the first place, because of what the tool was and how it was used. A system run inside a firm’s own environment, at counsel’s direction, under terms the firm controls, presents a different set of facts.

What follows for a firm is mostly procedural. Know which tools are actually in use. Keep client matters out of consumer services. And where AI touches privileged work, expect the next court to ask the same questions this one did: whose system, whose terms, whose direction.