AI Protective Orders: Controlling Unauthorized Exposure of Confidential Discovery to Generative AI Platforms

This CLE webinar will examine how protective orders are addressing the risk of opponents uploading confidential information produced in discovery to generative AI platforms that can then capture and disclose the material for training or to third parties. The panel will discuss this swiftly evolving but critical area of law and offer practical insights for effective and enforceable provisions for controlling the use of AI. The panel will also review guidance being provided by the courts.

Standard protective orders prohibit disclosure of confidential information outside specific protocols, include claw-back provisions, and require everyone to return/destroy all copies at the end of their involvement in the case. AI platforms typically disclaim all of these safeguards. 

Publicly available AI platforms keep all information submitted to it, use that information for training, and reserve the absolute right to share the information with third parties, although proprietary or closed AI programs may not be so broad. As a result, protective orders should now include provisions addressing the use of AI to analyze confidential materials produced in discovery, but these provisions are often disputed. Mechanisms for oversight and enforcement are crucial. Counsel should also be reviewing existing protective orders to determine the existence of gaps and possible fixes in pending cases. 

One renowned commentator has identified four emerging approaches for AI provisions in protective orders, ranging from a blanket prohibition to pre-use opportunity to object. Courts still grapple with whether AI should be considered more like a third party or just another tool or vendor, and what the consequences should be of inadvertent disclosure to an AI platform. Because the stakes are high.

Listen as our experienced panel offers best practices for ensuring effective AI provisions in protective orders.