Courts and litigants continue to grapple with the new frontier of artificial intelligence (“AI”).  One recent case in California demonstrates a new wrinkle in this evolving landscape—the use of AI to aggregate class claims.

Because class settlements bind absent class members who do not object or opt out, Rule 23 requires courts to carefully review and approve them as “fair, reasonable, and adequate.”  An important part of this inquiry is making sure class members are given adequate notice of the terms of the proposed settlement and their rights.  When class members are required to submit claims to access settlement benefits, parties often turn to professional claims administration companies to assist in providing notice and facilitating the claims process.  Under Rule 23, courts closely monitor the information that flows from class counsel and claims administrators to putative class members to make sure it complies with due process.

ClaimClam—a company that uses AI to aggregate and submit class claims—adds a troubling wrinkle to this process.  ClaimClam leverages its members’ personal information and AI to identify proposed settlements in which members might be able to participate, and then submit claims on their behalf.  In doing so, ClaimClam provides information to putative class members without oversight from class counsel, the claims administrator, or the Court.  This gives rise to significant concerns about whether putative class members are getting full and accurate information, and thus whether they are making informed choices about participation. 

In a recent class action involving Juul, ClaimClam attempted to submit tens of thousands of class claims and opt-outs “en masse.”  The claims administrator rejected these en masse claims in part because ClaimClam provided incorrect or misleading information to potential class members.  ClaimClam and its CEO objected, but on September 19, Northern District of California Judge William Orrick overruled ClaimClam’s objections and granted final approval of the class settlement.  In re Juul Labs, Inc., Mktg., Sales Practices, and Prod. Liab. Litig., 2023 WL 6205473 (N.D. Cal. Sept. 19, 2023).  In doing so, the Court determined that the settlement administrator had properly disallowed the en masse claims, reasoning that “[a]llowing en masse submissions by claims aggregators like ClaimClam raises real risks that Class Members will not receive accurate information regarding the scope of the class and the claims process.”  Id. at *9.  Further, en masse submissions impact the ability of the claims administrator to communicate directly with class members to, for example, weed out fraudulent claims.  Id.  To protect ClaimClam’s customers, the Court provided a mechanism for the ClaimClam claimants to individually submit claims.   

Putting aside the notice concerns that motivated the decision in In re Juul, AI claims aggregation also gives rise to significant concerns regarding the accuracy and validity of claims that are submitted.  If AI-assisted claims are inaccurate, qualifying class members may have their benefits diluted by participation of non-qualifying individuals.  And for certain claims-made settlement structures, defendants may end up paying out more than they otherwise should. 

Recent experience in mass tort litigation—where mass filing of unexamined claims is routine—provides reason for concern.  The MDL Subcommittee of the Advisory Committee on Civil Rules remarked in 2018 that “a significant number of [MDL] claimants ultimately (often at the settlement stage) turn out to have unsupportable claims,” including because “the claimant did not use the product involved” or “had not suffered the adverse consequence in suit.”  The Advisory Committee estimated that the “proportion of claims falling into this category . . . may be as high as 40% or 50%.”  Id.  It remarked that one reason offered to explain this phenomenon is “the effect of ‘1-800’ lawyers and ‘claims generators’ who support an atmosphere of ‘get a name, make a claim’ . . . in the expectation that there will be a settlement in the MDL transferee court in which they can get ‘inventory value’ for their claims.”  Id. at 143.  The incentives are much the same for class action claims generators like ClaimClam, which charges claimants a “service fee” of at least fifteen percent of any recovery.  ClaimClam, Authorized Agent Agreement § 2, https://drive.google.com/file/d/1JdDvAXaBvHnDvjW4tawExtgPoLMnK9pS/view.

We continue to monitor the impact of AI across litigation and especially in the class actions space.  We will highlight future AI-related developments in the blog.  Covington’s cross-disciplinary teams are ready to assist clients on a broad range of AI issues.