As companies and individuals increasingly embed AI tools in legal practice, courts are grappling with how to treat communications with, and information generated by, these tools. A primary concern is whether attorney-client privilege and work-product protections extend to AI-generated content or interactions with AI tools, depending on jurisdictional applications. The confidentiality of this information is directly impacted by these protections. Given the dynamic legal landscape surrounding AI, legal practitioners and their clients are advised to stay informed about evolving case law. It is crucial for anyone considering the use of AI tools in contexts such as litigation, regulated activities, compliance, or other legal endeavors to consult with counsel prior to implementation.
Sohyon Warner v. Gilbarco, Inc., Gilbarco, Inc. (d/b/a Gilbarco Veeder-Root), and Vontier Corporation
In this case, the United States District Court for the Eastern District of Michigan, Southern Division, addressed whether AI-use materials were protected work product and if using ChatGPT waived this protection. Occurring at the discovery-disputes stage, the court ruled against producing documents regarding ChatGPT use. It determined that the AI-use materials constituted the pro se plaintiff’s protected opinion work product, subject to heightened protections. Furthermore, the court found no waiver of work-product protection, reasoning that disclosure must be to an adversary or likely to reach an adversary. ChatGPT was classified as a tool, not a person, negating the conditions for waiver, even with potential administrators 'somewhere in the background'.
United States of America v. Bradley Heppner
The United States District Court for the Southern District of New York evaluated if AI-generated documents could be considered attorney-client communications or qualify as work product. In a post-indictment, pre-trial stage ruling, the court held that the documents were not privileged. This was based on three key reasons: the Claude AI tool is not an attorney, thus communications with it are not attorney-client; the communications were not confidential as Claude explicitly disclaims confidentiality and allows third-party sharing; and the defendant did not use Claude to obtain legal advice from the AI itself. Additionally, work-product protection was denied because the documents were not prepared 'at the behest of counsel' and did not reveal counsel’s strategy.
Morgan v. V2X, Inc.
During the discovery stage, the United States District Court for the District of Colorado considered whether FRCP Rule 26(b)(3) protects a pro se party’s AI-assisted litigation work under work-product protections. The court concluded that Rule 26(b)(3) can indeed apply to a pro se party’s AI-assisted mental impressions. It reasoned that using an AI tool does not automatically waive work-product protection simply because a third-party provider stores the data. Furthermore, AI interactions do not inherently compromise these protections because it is highly unlikely the information would fall into an adversary’s hands. However, the court mandated the plaintiff to disclose the name of the AI platform used, as revealing its name would not expose the plaintiff's mental impressions or case strategy, distinguishing it from the protected outputs.
