SDNY Court Rules AI-Generated Legal Documents Not Protected by Privilege

Court Decision on AI-Generated Legal Documents

On February 10, 2026, Judge Jed S. Rakoff of the U.S. District Court for the Southern District of New York ruled in U.S. v. Heppner that 31 documents generated using a third-party AI tool were not protected by either attorney-client privilege or the work product doctrine. This appears to be the first court decision addressing whether interactions with publicly accessible AI tools containing privileged information are themselves privileged.

Key Findings from the Ruling

Judge Rakoff rejected attorney-client privilege claims based on several specific findings:

• The defendant communicated with a third-party AI platform (Anthropic's Claude)
• The AI tool's privacy policy states the company collects data on both user inputs and tool outputs
• Collected data is used to train the tool
• The company reserves the right to disclose such data to third parties including government regulatory authorities

The court found there was a diminished expectation of confidentiality when using public AI tools. The ruling also declined work product protection because the defendant prepared the material without direction from counsel.

Case Background

Defendant Bradley Heppner was arrested on fraud charges on November 4, 2025. Before his arrest, he used Claude to:

• Run queries related to the government's investigation
• Prepare reports outlining defense strategy
• Outline arguments regarding facts and law he anticipated the government might charge

Defense counsel argued privilege because Heppner's inputs included information learned from counsel, the documents were created to obtain legal advice, and he shared them with counsel.

Legal Standards Applied

The court applied traditional privilege standards:

• Attorney-client privilege requires communications between client and attorney that are kept confidential for obtaining/providing legal advice
• Work product doctrine protects materials prepared by counsel or at counsel's direction in anticipation of litigation
• The party invoking privilege bears the burden of demonstrating it applies

This ruling suggests companies and litigants should carefully review AI tool terms, particularly regarding data use for training and disclosure permissions. While fact-specific, it establishes precedent for how courts may treat AI-generated legal materials.