On February 10, 2026, Judge Jed Rakoff of the Southern District of New York ruled from the bench that documents a criminal defendant created using an AI tool were not protected by attorney-client privilege or the work-product doctrine when those documents were later shared with defense counsel. A written memorandum explaining the court’s reasoning followed on February 17, 2026. As Judge Rakoff noted, the decision “appears to answer a question of first impression nationwide.”
In United States v. Heppner, the defendant used Claude, an AI chatbot developed by Anthropic, to run queries related to the government’s investigation before his arrest. He later shared approximately 31 AI-generated documents with his attorneys. The government moved to compel production of these documents, arguing they were neither privileged nor protected work product. Judge Rakoff agreed, stating: “I’m not seeing remotely any basis for any claim of attorney-client privilege.”
The court’s reasoning has significant implications for anyone using AI tools in connection with legal matters.
Attorney-client privilege & work-product doctrine legal framework
The attorney-client privilege protects confidential communications between a client and their attorney made for the purpose of obtaining or providing legal advice. The work-product doctrine “provides qualified protection for materials prepared by or at the behest of counsel in anticipation of litigation or for trial.” Whether these protections extend to AI-generated materials, however, has remained an open question—until now.
Attorney-client privilege limits
The defendant in Heppner argued the AI-generated documents should be privileged because they were “created for the purpose of obtaining legal advice from counsel” and “consolidate[d] what the defendant considered relevant so that he could discuss it with his lawyers.” In essence, he contended that the documents were akin to a memorandum prepared by an individual for discussion with their attorney.
The court rejected this argument on three independent grounds.
- First, communications with an AI tool are not communications with counsel. The privilege applies to communications between a client and an attorney. As the court explained, “[n]o such relationship exists, or could exist, between an AI user and a platform such as Claude.” Using an AI chatbot to research legal issues is not the same as consulting a lawyer.
- Second, the communications were not confidential. Anthropic’s privacy policy expressly permits disclosure of user data to “governmental regulatory authorities” and “third parties.” Under established law, the privilege is waived when communications are disclosed in a manner that negates the intent to maintain confidentiality. Users of consumer AI platforms have a “diminished privacy interest” in their conversations given these terms.
- Third, the communications were not made for the purpose of obtaining legal advice, at least not from Claude. The AI tool itself disclaims providing legal advice. While Heppner may have intended to share the outputs with his attorneys later, the relevant question is whether he sought legal advice from Claude. He did not, and could not, because Claude is not a lawyer.
Work-product doctrine limits
The defendant alternatively argued the documents qualified as protected work product because they were prepared in anticipation of litigation. He emphasized that the doctrine’s plain language protects materials prepared “by or for” a party, not exclusively materials prepared by attorneys. Under this reading, documents a client creates independently to assist with their own legal defense should qualify for protection.
The court rejected this argument, emphasizing that the doctrine’s purpose is “to preserve a zone of privacy in which a lawyer can prepare and develop legal theories and strategy with an eye towards litigation.” Second Circuit precedent stresses that the doctrine protects attorneys’ mental processes—their strategies, legal theories, and thought processes.
The court expressly disagreed with Shih v. Petal Card, Inc., a 2023 SDNY decision suggesting work product protection could apply to materials prepared by a party without attorney direction. Because the defendant “acted on his own” and the documents “did not disclose counsel’s strategy,” the work-product doctrine did not apply.
Broader implications
The ruling has implications that extend well beyond this case. For instance, the government’s argument that sharing data with a third-party platform defeats confidentiality has potentially sweeping implications for AI use. Many consumer AI tools have similar terms of service that permit disclosure to third parties or government authorities. Until the law develops further, the court’s reasoning likely applies to any publicly accessible AI platform with comparable privacy policies.
Until the law in this area develops further, we recommend the following practical steps.
1. CONSULT COUNSEL BEFORE USING AI FOR LITIGATION-RELATED MATTERS
If you anticipate litigation or are involved in a government investigation, speak with your attorney before using AI tools to research legal issues or organize relevant information. Your lawyer can help structure the engagement to maximize protection.
2. REVIEW THE AI PLATFORM’S TERMS OF SERVICE
Understand what you are agreeing to when you use AI tools. Privacy policies that permit disclosure to government authorities or third parties may affect the analysis of privilege.
3. TREAT AI COMMUNICATIONS AS POTENTIALLY DISCOVERABLE
If you are using a publicly accessible AI platform without counsel’s direction, assume that your interactions could be subject to discovery in litigation. Avoid entering sensitive information you do not want an adversary to see.
4. UPDATE CLIENT ENGAGEMENT PRACTICES
Lawyers should consider addressing AI use in engagement letters or during client onboarding. Proactive guidance can help clients avoid inadvertent waiver.
5. DOCUMENT PURPOSE AND TIMING
If you do use AI tools in anticipation of litigation, document that the materials were created “because of” the prospect of litigation. This contemporaneous record may support a work product claim if the issue arises.
6. CONSIDER ENTERPRISE AI SOLUTIONS
Some AI platforms offer enhanced confidentiality protections, data-handling commitments, or contractual terms that may better support expectations of privilege. Evaluate whether such solutions are appropriate for sensitive matters.
Preserving privilege post-Heppner
As Judge Rakoff observed, generative AI “presents a new frontier in the ongoing dialogue between technology and the law.” But, as his ruling demonstrates, AI’s novelty does not exempt it from longstanding legal principles, including those governing the attorney-client privilege and the work-product doctrine.
While Judge Rakoff’s memorandum provides the first judicial guidance on this issue, the legal landscape remains unsettled. Other courts may reach different conclusions, particularly on work product, where existing precedent supports broader protection for party-prepared materials. There is meaningful room for advocacy as this area of law develops.
We will continue to monitor this case and others addressing the intersection of AI tools and legal privilege.
For more information on the issues discussed in this alert, please contact your Nixon Peabody attorney or the authors of this alert.