Loose AI Prompts Sink Ships: How Heppner Shook the Legal Community
3.10.2026
Artificial intelligence has steadily begun to transform nearly every professional industry in the world, and this month Judge Jed S. Rakoff of the U.S. District Court for the Southern District of New York issued a decision in United States v. Heppner[1] that has made more waves in the legal community’s already persistent AI storm. The legal community has been abuzz discussing Judge Rakoff’s Feb. 10 oral ruling and his corresponding Feb. 17 written decision, which held that defendant Bradley Heppner’s AI queries and responses were not protected by the attorney-client privilege or work product doctrine. In his decision, Judge Rakoff noted that “the implications of AI for the law are only beginning to be explored.”[2] The Heppner decision is both a wake-up call and a warning to attorneys and their clients.
What Happened in Heppner?
In Heppner, a criminal defendant input materials received from his counsel regarding case strategy into the AI platform Claude. Heppner input this information without his counsel’s direction or request. During the execution of a search warrant, the government recovered multiple electronic devices containing Heppner’s AI prompts and outputs. Heppner’s criminal defense counsel argued that the government should not be able to view the content of Heppner’s AI queries because such content was protected by privilege and the attorney work product doctrine. The court disagreed, ruling that no protections attached to Heppner’s use of the AI platform.
Heppner’s use of Claude in connection with his case should not come as a surprise to the legal community. Nowadays, many people refer to generative AI tools such as ChatGPT or Claude as if they are sentient, or personal assistants, colloquially saying they “asked ChatGPT” a question or “talked to Claude.” In this respect, a client may have the false reassurance that what they input into these platforms is somehow unique compared to other search engines. However, these platforms are algorithms that review and dissect the data you input. Both the inputted information and the AI-generated responses are just as discoverable as a Google search.
Claude does not conceal how it learns from a user’s data. In fact, it would violate numerous data privacy laws if Claude’s methods of processing personal user data were hidden. Privacy policies, including the one on Claude’s website, openly inform users how their data is used. However, very few users actually read the fine print on these privacy policies, or even know these policies exist in the first place. It would probably surprise most people to learn that Claude’s privacy policy explicitly gives its parent company, Anthropic, the right to disclose a user’s data to third parties in connection with legal disputes and litigation.
The Aftermath of the Heppner Decision: What Should We Expect To See Next?
Heppner marks the beginning of what will surely be a series of rulings that contemplate the effect AI tools have on the cloak of confidentiality. For instance, would the court in Heppner have ruled differently if the defendant had used a closed AI enterprise system that kept the information confidential and inaccessible to the public? Courts may hold that the use of these AI enterprise systems, in collaboration and consultation with legal counsel, may be treated more akin to a third-party consultant, and thus privileged. Interestingly, the court in Heppner noted that “[h]ad counsel directed Heppner to use Claude, Claude might arguably be said to have functioned in a manner akin to a highly trained professional who may act as a lawyer’s agent within the protection of the attorney-client privilege.”[3]
In addition, we should expect to see more courts weighing in on the discoverability of AI “conversations.” In January this year, Judge Sidney H. Stein affirmed a motion to compel requiring broad discovery of AI training outputs in a copyright litigation.[4] The court noted that conversations with AI chatbots are distinguishable and weaker than the privacy interests and protections given to traditional person-to-person conversations, such as telephone conversations. We may see courts ordering vast discovery of AI chatbot conversations in future litigations.
What Firms and Practitioners Should Do Now
Attorneys should take immediate steps to protect both clients and law firms from inadvertent waiver of the attorney-client privilege. Specifically, attorneys should: (1) include robust disclaimers and warnings in engagement letters and email signatures alerting clients to the risks of using AI platforms in connection with their legal matters; (2) review the privacy policies of any AI tools used in the practice; and (3) advise clients to disable AI notetakers during confidential conference calls.
Elizabeth E. Schlissel is a partner at Falcon Rappaport & Berkman, where she practices in the Labor & Employment Law and Healthcare Practice Groups. She counsels management and HR on workplace policies, employment law compliance, and employment and separation agreements. She conducts workplace investigations and anti-harassment training, and represents employers in state and federal court and before government agencies in matters involving wage-and-hour violations, discrimination, harassment, retaliation, restrictive covenants, and disability accommodations. She also hosts the web series Employment Edge with Liz.
Daryl T. Caffarone is an associate at Falcon Rappaport & Berkman LLP in the Corporate & Securities, Intellectual Property, and Cannabis & Psychedelics Practice Groups. Her practice focuses on corporate transactions, including mergers and acquisitions, joint ventures, private placements, licensing, and corporate and partnership reorganizations. Daryl also advises on intellectual property compliance and registrations, and supports clients on regulatory and transactional matters in the cannabis industry.
Endnotes:
[1] 1:25-cr-00503(JSR), 2026 WL 436479 (S.D.N.Y. Feb. 17, 2026).
[2] Id. at *1.
[3] Id. at *3; cf. United States v. Adlman, 68 F.3d 1495, 1498-99 (2d Cir. 1995) (citing United States v. Kovel, 296 F.2d 918 (2d Cir. 1961)).
[4] In re OpenAI, Inc., Copyright Infringement Litig., 2026 U.S. Dist. LEXIS 1986, 2026 WL 21676, at *2 (S.D.N.Y. Jan. 5, 2026).
