AI in Court: Lessons from United States vs Heppner
The increasing ubiquity of AI technology in our lives has far reaching implications for corporations, individuals and the legal system. Many corporate leaders are increasingly using AI platforms as private brainstorming partners for strategy, technical specs, or sensitive data; individuals are turning to AI chatbots for personal legal and medical advice, so much so that the New York legislature is already moving to impose liability on AI companies that allow it; and the courts are wrestling with the implications of AI, as they are called upon to fit this new technology into existing legal frameworks.
The Southern District of New York’s recent decision in United States v. Heppner illustrates the kinds of unique issues the courts are now confronting.
In Heppner, the defendant used the generative AI platform Claude, operated by Anthropic, to inquire about legal strategy after receiving a grand jury subpoena and learning that he was the target of a criminal investigation. Notably, he made the decision to consult the popular AI platform without any instruction or suggestion from his attorneys that he do so. He memorialized his conversations with Claude in more than two dozen documents, which he saved to his computer’s hard drive. Those documents were later seized by the FBI when it was executing a search warrant at the defendant’s home. His attorneys argued that the documents were protected under both the attorney-client privilege and work product doctrine.
In what it said was a “case of first impression nationwide,” the Southern District (Rakoff, J.) held that the documents reflecting the defendant’s conversations with Claude were not protected by attorney-client privilege or the work product doctrine. The court’s reasoning applies equally to parties involved in criminal and civil proceedings.
Breaking Down United States v. Heppner
Finding that the defendant’s AI documents were not protected by attorney-client privilege, the court noted that they lacked “at least two, if not all three” requirements:
- First, Claude is an AI tool and not an attorney; communications with an AI platform simply do not occur in the context of a relationship with a licensed legal professional.
- Second, Claude’s AI data privacy policy provided “no reasonable expectation of confidentiality” in the defendant’s communications with the platform. The court noted that under Anthropic’s written privacy policy, users consent to collection of both their own “inputs” and Claude’s “outputs,” with Anthropic reserving the right to use such data for training and to disclose it to third parties, including governmental authorities.
- Finally, while Heppner’s counsel asserted the defendant had used the platform for the “express purpose of talking to counsel,” the court noted that he did not do so at the direction or suggestion of counsel. His intention to share the documents after the fact did not change that conclusion, the court finding that because the AI documents would not be privileged had they remained in the defendant’s hands, they did not acquire protection merely because they were to transferred to his legal counsel. The court also used this point when ruling that the documents were not protected by the work product doctrine, which protects only materials prepared by or at the behest of counsel in anticipation of litigation or for trial.
Heppner does not create a blanket exception excluding all AI-generated documents from privilege, but it does provide insight into how the courts can be expected to apply existing doctrines and rules of evidence to the products of this new technology.
Corporate Implications: High Stakes Across Industries
Heppner raises several important questions about AI, the rules of evidence, and how corporations can proactively protect themselves:
- How can corporations utilize AI technology while limiting risk and liability?
- How can litigation teams use AI?
There is significant risk to corporations and their counsel if they feed sensitive data (e.g., financial or other confidential/proprietary information) into a public AI platform where, according to its own terms of service, no expectation of privacy or confidentiality exists. Internal business data shared by employees with free, publicly available versions of AI tools represents a significant cybersecurity and litigation vulnerability.
Corporations that would benefit from AI technology must implement enterprise-level security, infrastructure, and training guided by strict AI usage policies. Companies must also conduct comprehensive audits to identify whether and where sensitive data is already being shared on third-party platforms of any kind.
AI tools and their governance should be vetted by legal counsel to ensure data privacy is properly protected, limiting potential liability and the exposure of confidential data.
Bridging the Gap Between Technology and the Law
Technology sometimes moves faster than the law, but Heppner shows that existing rules of evidence still apply. The case also demonstrates that while technology like AI can be a useful and even valuable aid, the law cannot be expected to accommodate its use in ways that erode long-standing principles governing the attorney-client relationship and the role of attorneys as licensed fiduciaries providing legal counsel and advice.
Bleakley Platt & Schmidt’s Information Technology and Cybersecurity practice group can help companies develop a comprehensive AI governance strategy that protects their data and their business. Call us today to set up a consultation: 914-949-2700.