New York Court: AI-Generated Documents Not Protected by Privilege
A federal court in New York has issued a first-of-its-kind ruling: documents created by a client using generative AI are not protected by attorney-client privilege. This decision serves as a critical warning for how legal teams and their clients interact with artificial intelligence.
While the ruling was specific to the facts of this case, its implications are broad. It shows that using AI tools can create discoverable evidence, potentially usable against a corporation in court.
Case Background
The case involved Bradley Heppner, who faced charges of securities and wire fraud. After receiving a grand jury subpoena and hiring a lawyer, Heppner used the AI tool Claude to generate over 30 documents exploring defense strategies.
He then shared these documents with his attorney. Crucially, his lawyer never asked him to use an AI tool for any purpose. Heppner did it entirely on his own.
His attorney argued the documents were privileged client-counsel communications. The government countered that sending non-privileged documents to a lawyer doesn't magically make them privileged.
The Court's Ruling on AI and Privilege
Attorney-client privilege requires communications to be (1) between a client and attorney, (2) confidential, and (3) for the purpose of getting legal advice. The U.S. District Court for the Southern District of New York found Heppner's AI documents failed on all three points.
First, by sharing information with a third-party AI platform, Heppner was not communicating with his attorney. There was no direct client-counsel interaction in the creation of the documents.
Second, he had no reasonable expectation of privacy. Claude's user policy states that inputs may be used for AI training and disclosed in litigation. This lack of confidentiality breaks the privilege.
Finally, since his lawyer never directed him to use AI, the act was not for the purpose of obtaining legal advice from counsel. It was independent research.
Key Takeaways for Legal Professionals
This ruling provides clear action items for anyone in the legal field. Ignoring them creates unnecessary risk.
- Assume AI Inputs Are Not Confidential: Treat any information entered into a public AI platform as public. Even if you later share the output with counsel, the initial act of creation is not a protected communication.
- Privilege Can Be Destroyed: Entering existing privileged information into an AI platform to summarize or analyze it can waive that privilege. The information has been disclosed to a third party.
- Read the Fine Print: Carefully review the terms of service for any AI tool. If the policy states your data can be used or disclosed, you cannot claim a reasonable expectation of privacy.
- Update Corporate Policies: Companies must create and enforce clear policies on employee AI use. Educating your teams on the proper and improper use of AI for Legal work is essential to prevent employees from accidentally generating discoverable evidence.
Your membership also unlocks:
