Cigna PxDx Batch Rubber-Stamp Denials
Cigna's PxDx system was used to deny 300,000+ medical claims over about two months in 2022, with reviewing doctors averaging a reported 1.2 seconds per denial.
What happened
According to a 2023 ProPublica investigation and subsequent lawsuits, Cigna's automated PxDx claims system generated denial recommendations for medical benefits requests, and company-employed doctors then signed off on these denials in batches. ProPublica reported that over a period of about two months in 2022, Cigna doctors denied more than 300,000 requests, spending an average of 1.2 seconds on each. Plaintiffs allege the doctors did not open or individually review the claim files before denying them; Cigna disputes the characterization and, in litigation, has contested whether the named plaintiffs were even subject to PxDx review. In March 2025, a federal court in the Eastern District of California issued a mixed ruling on Cigna's motion to dismiss: it allowed some claims to proceed (including a breach of fiduciary duty claim under ERISA and a California Unfair Competition Law claim) but dismissed the wrongful benefit denial claims for insufficient pleading and found the three named plaintiffs had not shown they were subjected to PxDx review, granting leave to amend.
What the agent did
Human doctors, not the algorithm, signed off on the denials: reviewers approved 300,000+ PxDx-generated denials in batches over about two months in 2022, at a reported average of 1.2 seconds per denial. Plaintiffs allege the reviewers did not read individual claim files before finalizing the denials. The algorithm produced the denial recommendations; the commitment was a human action.
The irreversible effect
More than 300,000 claims were denied over about two months in 2022, and most were not appealed; affected patients lost access to benefits unless they filed appeals or pursued litigation. Whether those denials were wrongful is alleged in the lawsuits, not an adjudicated finding. The practice was surfaced by a 2023 ProPublica investigation and has led to class action litigation, some of which a court allowed to proceed in March 2025.
Root cause
Absence of mandatory approval gate requiring reviewers to spend measurable time and individually document their review of each claim before committing an irreversible denial; batch approval workflows allowed decisions to be finalized without meaningful human oversight; no audit trail recorded decision time, reviewer rationale, or per-claim review before commitment.
How a maker-checker control would have refused it
high_risk_requires_gate—if the denial commitment skill were designated as high-risk, any attempt to execute it would be refused on the proxy with the message: 'skill cannot run through the proxy; run it in a governed flow with a preceding approval gate'. This would force all denials through an approval gate that records the reviewer's identity, elapsed time between claim presentation and decision, and individual claim review, making batch rubber-stamp patterns visible and attributable in the audit trail rather than hidden in internal metrics.
Runnable reproduction
This incident ships as a runnable scenario in the open-source repository. Point the enforcement engine at the policy and watch the action get refused, with the refusal written to a signed audit record.
examples/cigna-pxdx-batch-rubber-stamp-denials
Accuracy and corrections
This entry describes a publicly reported incident and is compiled from the primary sources listed above. Where an account is a legal allegation rather than an established finding, the entry labels it as such. Summaries can still contain errors. If you can document a correction, email hello@makerchecker.ai and we will review and correct it, with the change noted, within 14 days.
See it for yourself
Reading is one thing. Watch it block an agent.
One command starts the demo: an agent stopped from signing off its own work, and the signed evidence file an inspector can check for themselves.
Designed against the rules your auditors already enforce.