United States v. Heppner: What Every Litigator Needs to Know About AI and Privilege
I don’t practice law anymore, but I still nerd out on legal concepts. It’s probably why I decided to remain in the industry long after my billable hour days were over. I find the increased use of AI fascinating in general, but the implications in the legal industry are worth paying extra attention to.
While there are a lot of ways AI is reshaping actual attorney work product, I read a case the other day that raised a point I hadn’t thought of but really opened my eyes. It brings to light considerations surrounding attorney-client privilege and the work product doctrine which are both cornerstones of legal practice.
On February 10, 2026, Judge Jed Rakoff of the Southern District of New York came across a question of first impression at the federal level. Does the use of a third-party AI platform preclude litigants and litigators from the protection of attorney-client privilege or under the work product doctrine? In at least one situation it does. In United States v. Heppner the court ordered 31 documents generated by the defendant using a consumer version of Claude to be turned over to the prosecution, finding that these AI-generated documents did not fall under attorney-client privilege and were not protected by the work product doctrine. A defendant who used AI to organize his legal strategy and then shared that work with his attorneys got no protection at all. The documents that he thought were private became evidence against him.
The reasoning was straightforward, although it created some gray areas as well. First, attorney-client privilege requires confidentiality, and the moment Heppner typed sensitive information into a consumer AI tool operated by a third party, he effectively shared it outside the attorney-client relationship precluding protection. Additionally, the AI drafted documents were not attorney work product because they were not from an attorney.
This significant ruling creates a lot of questions and new considerations for litigators when advising their clients. Attorneys will need to advise clients explicitly on the use or non-use of AI, and law firms as well as in-house legal counsel will need to think carefully about what the company’s policies will be.
We are still unclear as to whether an attorney directed use of AI would allow for protection or whether an attorney’s own inputs would be considered work product, but the court appeared to leave some fact-dependent protections on the table. Clearly, that gap between what clients are doing every day and what attorneys have told them about it is where the exposure lives.
The Heppner case is one data point amidst these new, rising questions, but it won’t be the last. AI is going to be an interesting ride, folks. I think we all need to buckle up.
Do you agree with Judge Rakoff’s decision? Do you think the substance or impetus for the use of the AI program should make a difference with respect to privilege? Should enterprise accounts be treated differently than free-to-use services?
Carrie Marcantonio is the owner of CB Consulting Group, LLC, an independent legal recruiting firm specializing in attorney placement.
Citation: United States v. Heppner, No. 25-cr-00503-JSR (S.D.N.Y. Feb. 17, 2026), Dkt. No. 27.