AI agents could upend class action settlement claims

Artificial intelligence may soon do more than help lawyers research cases and draft briefs. AI agents could begin finding class action settlements for consumers and filing claims on their behalf, potentially changing the economics of a system in which only a fraction of eligible class members sometimes collect money.

Class action lawyers say personal AI agents could search users' email and the web for settlement notices, determine whether they may qualify and help complete claim forms. At the same time, class counsel and settlement administrators could use AI to find and communicate with class members.

"I think it's a super exciting time," said Jay Edelson, founder of plaintiffs' firm Edelson PC. "People's own AI agents are going to be able to do searches, whether it's on their email or also searches on the web, and just find out about class actions where they can participate in much easier."

Edelson said class action lawyers and claims administrators are also likely to deploy the technology themselves to reach class members and guide them through the claims process. He predicted agentic AI will be a "game changer on both sides."

Nathaniel E. Haas, an attorney with Watstein Terepka LLP, said AI could build on technological changes that have already made settlements easier for consumers to navigate. Electronic payments, for example, have replaced the process of submitting an address and waiting for a check in some settlements with direct deposits and payments through services such as Venmo.

AI agents could help at an even earlier stage by finding notices that consumers otherwise overlook, Haas said.

"There's just such a high rate of people who never see these," Haas said of notices that wind up in spam folders. "The goal of class action settlements is to actually reach these people, and it seems that AI and the agentic process may make that a bit easier, which is a good thing."

How much higher participation would change settlement economics depends on the structure of the deal, according to Gretchen L. Jankowski of Buchanan Ingersoll & Rooney PC.

"Under a claims-made structure -- where the defendant's liability is primarily determined by the number of filed and qualifying claims -- a narrower gap would likely cause defendants to face a much larger total payout than is traditionally seen," Jankowski said.

The effect could run in the opposite direction for individual claimants when defendants agree to pay a fixed common fund.

"Under the common fund structure -- where the total payout is fixed regardless of the number of successful claimants -- a narrower gap would tend to negatively impact the members of the class," Jankowski said. More successful claimants would be competing for shares of the same pool of money.

Edelson predicts the consequences could be more disruptive. If AI makes it easy for far more class members to participate, he said, some defendants would become less willing to settle because the amount they actually have to pay consumers could increase dramatically.

"It's actually going to make class action settlements harder to reach," Edelson said. "We're going to see more cases go to trial."

He also said he believes that higher participation could expose settlements that appear substantial on paper but deliver relatively little value when measured across the entire class.

Edelson offered the hypothetical of a claim worth $1,000. A settlement designed around a 1% claims rate might pay each claimant $200, he said. If 10% of the class suddenly participates, that could fall to $20; if everyone participates, it could fall to $2.

Edelson sharply criticized what he described as a portion of the plaintiffs' bar that benefits from low claims rates, arguing some plaintiffs' lawyers and defendants have incentives to structure settlements around the expectation that few class members will participate.

If AI eliminates that expectation, he said, defendants facing substantially greater actual payouts may instead decide to litigate.

Christopher Chorba, a class action defense attorney at Gibson Dunn & Crutcher LLP, said he was skeptical that AI will cause anything approaching that transformation.

"I think we'd have to be careful and skeptical of any significant uptick in percentages," Chorba said. "Through several dozen years of experience, those just aren't achievable."

Chorba said low participation does not necessarily mean consumers were unable to find or complete claims. In some cases, he said, consumers may decide the available benefit is not worth pursuing or dispute whether the defendant did anything wrong.

"If, for example, a settlement were having a low single-digit percentage claims rate," and participation suddenly rose to the mid- or high-double digits, Chorba said, "I would be very, very skeptical that that increase is attributable to legitimate claimants, and I would strongly suspect that that would be increased fraud."

That concern points to another way AI could change settlement administration: making it easier for both legitimate class members and fraudsters to submit claims.

Chorba said the danger is particularly acute in cases involving retail goods, food or other products where defendants lack records identifying every purchaser.

"Those are the ones that are really ripe for fraud," he said, because there may be no "master class list" against which administrators can easily check whether someone actually belongs in the class. He said parties, courts and administrators will have to remain vigilant as AI becomes another tool people can use to locate settlements.

Jankowski said courts and administrators are already encountering mass claims generated with technological assistance.

She pointed to the Juul Labs litigation, where she said an administrator rejected tens of thousands of claims submitted "en masse" through ClaimClam, a company that used AI to collect and submit claims while charging claimants 15% of their recoveries. In the Novartis and Par Antitrust Litigation, Jankowski said the administrator rejected nearly 80% of submitted claims as fraudulent.

"The increasing prevalence of AI agents in the claims submission process requires class counsel to be more diligent in choosing claims administrators that have robust systems in place to detect fraudulent and duplicative submissions," Jankowski said.

Chorba said the answer should not be to presume that a claim is fraudulent simply because AI helped produce it. Administrators already investigate suspicious claims, he said, and can require additional verification when patterns suggest fraud.

"The issue isn't necessarily the source," Chorba said. "The issue is fraud detection and prevention."

Adam E. Polk, a plaintiffs' attorney at Girard Sharp LLP, similarly said AI-assisted claims may lead to additional verification, but administrators may be able to use technology to scrutinize those claims without putting new obstacles in front of legitimate class members.

Polk also cautioned against taking consumers entirely out of the claims process.

"We would be hesitant to adopt any technology that completely automates this process out of due process concerns," Polk said. Class members should still read settlement notices, understand their rights and review claims submitted for them, he said.

"If AI agents are able to increase class member participation without diminishing class members' opportunity to become informed of the settlement's terms and their corresponding rights," Polk said, "I think this could be a net positive."

That leaves the class action system facing two competing possibilities. AI could make it easier for eligible consumers to receive money that settlements were intended to deliver, while simultaneously making it easier to generate fraudulent claims at unprecedented scale.

For Haas, the underlying objective remains straightforward.

"The big concern is they get notified of these class settlements in their spam folder," he said. If AI helps more legitimate class members find those notices and collect their share, he added, "that's a good thing."