Attorney-Client Privilege Meets Generative AI

Here is my recent Daily Record column. My past Daily Record articles can be accessed here.

****

Attorney-Client Privilege Meets Generative AI

Last week, a ruling was handed down that could have wide-ranging implications for lawyers using digital technologies like email and the cloud. The holding—that a legal client’s conversations with the generative AI tool, Claude, were not protected by attorney-client privilege—is not, in and of itself, problematic. 

However, Southern District of New York Judge Jed S. Rakoff subsequently issued https://storage.courtlistener.com/recap/gov.uscourts.nysd.652138/gov.uscourts.nysd.652138.27.0.pdf)https://storage.courtlistener.com/recap/gov.uscourts.nysd.652138/gov.uscourts.nysd.652138.27.0.pdf earlier this week clarifying his ruling and reaching conclusions that, if applied broadly, could preclude the use of technology long believed to be protected conduits for the exchange and storage of confidential and privileged information.

In this case, the defendant, of his own volition, accessed the free version of the generative AI tool, Claude, to draft defense strategy materials, which he then provided to his lawyer. Because no attorney was involved when he used Claude, the Court concluded that no attorney-client relationship existed between the defendant and the AI tool, and therefore, there was no privilege to waive.

Next, the Court determined that because Anthropic’s privacy policy permits model training and the collection of user inputs and AI outputs, along with possible disclosure to regulators, there was no reasonable expectation of confidentiality. According to the Court, sharing information with Claude would waive any applicable privilege, since it’s no different from sharing it with any other third party. Furthermore, the Court explained that privilege did not subsequently attach when the defendant provided those AI-drafted documents to his attorney since “non-privileged communications are not somehow alchemically changed into privileged ones upon being shared with counsel.”

Early on in the Order, the judge addressed and rejected the argument that inputs to AI tools are akin to data stored in SaaS software and thus are not "communications" at all. He concluded that privilege requires the existence of a “trusting human relationship,” which does not exist “between an AI user and a platform such as Claude.”

That rationale arguably conflicts with the longstanding treatment of digital tools. Courts have consistently held that using third-party software, including email and cloud services, does not by itself waive the attorney-client privilege, even when providers have technical access to the data. Instead, precedent holds that email and cloud providers are simply service vendors or agents that facilitate legal representation, and entering confidential information into these tools does not impact privilege. 

I can't imagine this determination will withstand the test of time. Taken to its logical conclusion, it could require that any information analyzed or stored in cloud tools, including email, would be deemed to waive privilege. 

The ruling strikes me as one of those knee-jerk reactions to emerging technology that often occurs in the earlier stages of its adoption. When a new, unfamiliar tool threatens familiar workflows and processes, the default response is distrust.

AI has become so ubiquitous, so quickly, that it’s unsettling, even for those of us who embrace technology and enjoy experimenting with new tools. Reticence and suspicion are normal, expected reactions.

However, given the rapid pace of change, attitudes will need to shift faster than they have in the past. AI technology is evolving quickly and reshaping the business landscape as it improves. Our profession will need to keep pace, and rulings like this one create roadblocks to adoption. 

What is troubling about this Order is the suggestion that AI tools are inherently different from the digital infrastructure lawyers have relied on for decades. By framing an interaction with Claude as disclosure to a third party because no “trusting human relationship” exists, the opinion hints at an AI exception.If that distinction holds, it could disrupt settled understandings about how confidential information is handled in modern practice. The better approach would be to avoid treating AI differently from other technologies and instead apply existing privilege principles consistently by focusing on the role the technology plays, who directs its use, and whether reasonable steps were taken to preserve confidentiality. Doing so would enable the flexible application of privilege principles to future technology advancements.

Nicole Black is a Rochester, New York attorney, author, journalist, and Principal Legal Insight Strategist at 8am, the team behind MyCase, LawPay, CasePeer, and DocketWise.She is the nationally-recognized author of "Cloud Computing for Lawyers" (2012) and co-authors "Social Media for Lawyers: The Next Frontier" (2010), both published by the American Bar Association. She also co-authors "Criminal Law in New York," a Thomson Reuters treatise. She writes regular columns for Above the Law, ABA Journal, and The Daily Record, has authored hundreds of articles for other publications, and regularly speaks at conferences regarding the intersection of law and emerging technologies. She is an ABA Legal Rebel, and is listed on the Fastcase 50 and ABA LTRC Women in Legal Tech. She can be contacted at niki.black@mycase.com.