Companies are increasingly turning to artificial intelligence ("AI") platforms to obtain legal information. In the intellectual property context, common uses include freedom-to-operate searches, claims drafting, and assessments of potential patent infringement exposure. A recent decision from the U.S. District Court of the Southern District of New York calls into question whether documents generated through certain AI platforms are entitled to protection under the attorney-client privilege and work product doctrines. In United States v. Bradley Heppner, Judge Jed S. Rakoff ruled in a criminal proceeding that documents generated through an AI platform were not entitled to protection under either the attorney-client privilege or the work product doctrine. In that case, the defendant is alleged to have used a consumer version of Anthropic’s Claude AI to generate documents related to his defense. Heppner’s attorney claimed that those documents were privileged and the Government subsequently filed a motion for a ruling that the documents were not privileged.
In granting the government’s motion on February 10, 2026, Judge Rakoff noted that the attorney-client privilege attaches to communications for legal advice between a client and their attorney that are intended to be, and are kept, confidential. Judge Rakoff determined that the AI-generated documents failed to meet the criteria for establishing privilege. First, the Judge pointed out that the AI documents were not communications with counsel, and the AI agent cannot be construed as legal counsel. Second, the communications in the AI-generated documents were not confidential because the communications were with a third party whose privacy policy explicitly states that users consent to Anthropic’s use of inputs and outputs for various purposes. Finally, the Judge ruled that Heppner’s use of AI was not for the purpose of obtaining legal advice. Heppner’s attorney suggested that Heppner was using AI for the purpose of communicating with counsel, but the Judge noted that Heppner did not do so at the direction of counsel (which, if he had, might still not have been sufficient to be considered attorney-client communication).
Key Takeaways:
- The use of AI to obtain legal analysis or advice might not be covered under the attorney-client privilege and work product doctrines. That means that there is a risk that AI-generated documents might not be protected from disclosure to an opposing party in litigation. Even the AI queries themselves may not be privileged and may later be construed as damaging admissions.
- The decision in US v. Heppner is a criminal case and is a decision of a single United States District Court. It appears to be the first decision regarding the use of AI to obtain legal advice, and it is possible that other District Courts or a Court of Appeal could render a different decision.
- The Heppner case does not specifically address the use of AI in the context of IP litigation, but it is reasonable to assume that the principles applied in Heppner would apply to any privilege contention. However, it is possible that another District Court or an appellate court would rule differently in the context of patent issues.
- Judge Rakoff noted that the defendant utilized a commercial version of AI. It is possible that an enterprise version would support a stronger expectation of privacy by the user.
- It is unclear whether the use of AI at the direction of counsel would result in a different decision.
What You Should Do:
Companies should exercise caution in their use of AI to generate legal advice, documents, and the like. It is best to consult with counsel prior to undertaking the risk of using AI for legal advice.
We will post further commentary as this area of the law develops.
