Cigna's CEO Told Congress AI Never Denies

Introduction

In January, David Cordani sat before the House Ways and Means Committee and told lawmakers that at Cigna, AI is "never used for a denial." A company spokesperson backed the point, saying the insurer's claims process "is not powered by AI," per KFF Health News. Cordani has run Cigna since 2009, and the company covered or administered health plans for roughly 18 million people as of ProPublica's 2023 reporting.

Ten months earlier, a federal judge had already read Cigna's own defense of the system those claims run through and written it into an order. Here is how the judge summarized Cigna's position: an algorithm makes the medical-necessity call, "so long as a medical director pushes the button." Cigna's internal documents, quoted in that same order, clock the button-push at an average of 1.2 seconds per claim, across more than 300,000 denials in a single two-month stretch in 2022.

The 1.2-Second Denial

If Cigna administers your health plan, that batch process is what sits between a flagged claim and a denial. A federal court has already found that the way Cigna ran it could break the plan's own promise that a medical director, an actual human, would decide whether your care was necessary. That is the whole tension inside Cordani's sentence to Congress. It holds up only if a person clicking "approve" on an algorithm's batch decision counts as that person making the decision, and a judge who looked at that exact argument called it an abuse of discretion.

The number that started all of this surfaced in March 2023, when ProPublica and The Capitol Forum published internal Cigna records showing how a system called PxDx worked. PxDx (procedure-to-diagnosis) flags claims where the billed procedure doesn't line up with an approved diagnosis code, then routes them for denial in bulk. Reporters Patrick Rucker, Maya Miller, and David Armstrong found that over two months in 2022, Cigna doctors denied more than 300,000 requests for payment this way, at an average of 1.2 seconds each. One medical director, Dr. Cheryl Dopke, denied roughly 60,000 claims in a single month.

That average is the part I keep coming back to. It isn't enough time to open a claim file, never mind read one.

What Judge Drozd Actually Held

The case is Kisting-Leung v. Cigna, filed in federal court in Sacramento in July 2023. On March 30, 2025, U.S. District Judge Dale Drozd ruled on Cigna's motion to dismiss, and this is the part to be precise about, because it happened over a year ago, not last week. He didn't rule Cigna liable, only that the case could go forward to discovery.

What he let through is narrow and specific. Cigna's plan documents promised that medical-necessity determinations would be made "by a Medical Director." Cigna's defense was that this promise was met because a medical director technically signed off on the algorithm's batch denials. Drozd rejected that reading in a line Cigna's own lawyers now have to work around: letting "an algorithm to make the decision so long as a medical director pushes the button" conflicts with the plain language of the plan and "constitutes an abuse of discretion."

It is not a clean win for the patients. Three of the six named plaintiffs, including the one whose name is on the case caption, were dropped from the algorithm claim after Cigna produced a sworn declaration that their specific claims never ran through PxDx. The straightforward "you owe me this benefit" claim was dismissed for everyone. What survives is the narrower fiduciary-duty theory, that Cigna's process itself violated the plan's terms, and it survives for three plaintiffs: Samantha Dababneh, Randall Rentsch, and Abdulhussein Abbas. No class has ever been certified, so this remains a putative class action of six people and their lawyers.

Here is why it is worth writing about now rather than in the spring of 2025: the case didn't end. Plaintiffs chose not to refile an amended complaint and pressed ahead on the surviving claim, Cigna answered in May 2025, and the docket tracked by Georgetown's Health Care Litigation Tracker shows fresh scheduling orders dated April 29 and May 1, 2026, with the case listed as ongoing. So when Cordani told Congress this January that AI never denies a claim, his company was, at that moment, defending in federal court the algorithm that does.

Cigna's own investors are being told to take it seriously. In the company's 2026 proxy statement, filed with the SEC in March, a shareholder proposal cites the PxDx ruling by name as a governance risk and quotes the "abuse of discretion" finding back to the board. The same proposal notes Cigna's stock fell from $340 in 2022 to $279 in late 2025. Cigna's board recommended shareholders vote against it. Bloomberg Law flagged a quieter problem: the notice Cigna sends when it uses PxDx is itself proof the technology was used, and that notice could cut against the company as discovery moves along.

Who Benefits

Cigna, first and most simply, in dollars. The complaint the court quoted says it directly: "the PXDX system saves Cigna money by allowing it to deny claims it previously paid." Alan Muney, Cigna's former chief medical officer, who helped build PxDx, told ProPublica the system "has undoubtedly saved billions of dollars." Every claim an algorithm bounces in 1.2 seconds is a claim Cigna doesn't pay and doesn't have to put a doctor's time into reading.

The subtler benefit is the one Cordani collected in that hearing room. By drawing the line at the word "AI" (insisting PxDx "is not powered by AI"), Cigna gets to answer a question that nobody's actual objection depends on. The March 2025 order doesn't turn on whether PxDx meets a technical definition of artificial intelligence. It turns on whether any automated system, of any kind, can make a call the contract reserved for a human. So Cordani can say "AI is never used for a denial" and have it hold up narrowly, while the practice the judge actually flagged goes unmentioned.

Arguing About the Label

Cigna's strongest defense skips the word "AI" entirely: PxDx, the company says, is post-service review. It runs after you have already gotten the care, so it isn't prior authorization and no treatment is blocked. The company says the flagged procedures are around 50 common, low-cost codes that mostly get approved, and that in-network patients shouldn't see a bill for a PxDx denial anyway.

Some of that lands. A post-service denial isn't a gate in front of your surgery. But it still decides whether Cigna pays, and "we didn't deny the care, only the payment" is thinner comfort when you are the one holding the bill. Post-service denials are how people end up balance-billed by an out-of-network provider, or grinding through an appeal for a procedure they already had. The distinction Cigna draws protects Cigna's framing, not your wallet, when a claim gets bounced in a second and change.

There is a reason the state-law piece of this case matters more now than it did in 2023. The surviving California claim rests on a state statute requiring that a licensed professional, not software, make medical-necessity calls. In December 2025, the White House signed an executive order directing the Justice Department to stand up an "AI Litigation Task Force" whose job is to challenge state AI laws. Cigna isn't named in it, and I won't pretend the order was written for this case. But the category of law it targets is the same category holding up the one claim against Cigna that a California court can hear. If those state laws fall, the legal foothold this fight stands on gets narrower for the next patient.

The Bottom Line

Strip away the semantics about the word "AI" and here is what remains: a system that denied more than 300,000 claims in two months at 1.2 seconds each against a plan that promised a doctor would make those calls, and a judge who found the gap between them serious enough to keep the case alive. Cordani told Congress none of it is AI, and he may well be right, which was never the point.

The case keeps moving through the court in Sacramento while Cigna tells its shareholders this litigation is a governance risk worth disclosing, and tells Congress the algorithm at the center of it barely exists. What I would watch is whether "a medical director pushed the button" still counts as a doctor deciding your claim if this ever reaches a jury, and how many claims run through that button while everyone keeps arguing about what to call it.