What Heppner and Warner Say, and What They Do Not

Two federal court decisions issued one week apart in February 2026 reached different conclusions about artificial intelligence and legal privilege. On 10 February 2026, Judge Jed S. Rakoff of the U.S. District Court for the Southern District of New York delivered a bench ruling in United States v. Heppner, followed by a written memorandum on 17 February 2026. He held that written exchanges between the defendant, Bradley Heppner, and a publicly available generative AI platform were protected by neither attorney-client privilege nor the work product doctrine. The court therefore allowed the government to inspect the logs.

Also on 10 February 2026, the U.S. District Court for the Eastern District of Michigan reached a different conclusion on work product in Warner v. Gilbarco, Inc. The court held that a pro se plaintiff's use of generative AI tools to prepare her litigation did not waive work product protection. It reasoned that AI platforms are tools, not persons, and that submitting information to a tool does not constitute disclosure to an opposing party.

This article is for informational purposes and is not legal advice. It examines how the facts led to two different rulings.

The outcome in Heppner turned on four narrow factual findings:

First, Heppner used a public consumer version of the software, not an enterprise or proprietary service.

Second, no attorney instructed him to use the platform. Neither Heppner nor the software acted as an agent of legal counsel.

Third, the platform's published privacy policy disclosed its data collection practices and rights to share data with third parties. That language undercut any claim to a reasonable expectation of confidentiality.

Fourth, legal privilege protects confidential communications with a licensed attorney. Software is not an attorney.

Legal commentators noted that Heppner leaves room for different outcomes under different facts. The analysis may change when an attorney directs the use of the software, the software is an approved enterprise tool, or contractual terms create a reasonable expectation of confidentiality.

Heppner and Warner show that the law is unsettled. Several major law firms and the International Bar Association describe U.S. federal courts as split on this question. A New York state court also addressed generative AI and privilege in July 2026. The governing legal standards continue to change.

Practitioners should evaluate the specific terms under which a system processes data. Three questions matter:

Where does the material go during processing?

Who has the right or technical ability to read or review the input?

Does the vendor's policy claim ownership, training rights, or the right to share user data with third parties?

Privilege analysis depends on the details. Storage, transmission, and contract terms can determine the result.