In a January 2026 congressional hearing, Representative Robin Kelly pressed UnitedHealth Group CEO Stephen Hemsley over his company’s use of artificial intelligence in the prior authorization and claims process, following years of reports that the insurer’s AI-driven denial rates have run far higher than industry norms in certain categories of care. According to Kelly’s office, UnitedHealth’s denial rate in some cases has run 16 times higher than typical, intensifying scrutiny of how insurers deploy algorithmic decision-making in coverage determinations that directly affect patient access to care.
Hemsley reportedly could not commit to specific safe-AI practices when questioned directly, according to reporting from Kelly’s office, a response that lawmakers and patient advocates characterized as evasive given the stakes involved. The hearing is the latest escalation in a fight that has been building since a 2023 lawsuit accused UnitedHealth of improperly denying claims using an algorithm called nH Predict, which critics allege carries an error rate as high as 90% when its denials are appealed and overturned.
The Legal Battle Beyond UnitedHealth
UnitedHealth is not alone in facing litigation over algorithmic claims denial. Cigna Corp. is defending against lawsuits tied to its PxDx algorithm, which according to court filings denied more than 300,000 payment requests over a two-month span in 2022, with Cigna physicians reportedly spending an average of just 1.2 seconds reviewing each flagged claim before denying it. As of 2026, judges in multiple jurisdictions are still weighing motions to dismiss in these cases, with courts wrestling over a novel legal question: whether processing claims through an AI system without meaningful individualized human review breaches the “reasonable investigation” standard insurers promise policyholders under state bad-faith insurance laws.
How Widespread Is AI in Claims Processing Now
This isn’t a fringe practice. Industry surveys found that 71% of health insurers reported using AI for utilization management — the broad category covering prior authorization and concurrent care review — as of 2025, meaning the legal and regulatory questions raised in these cases have implications far beyond any single insurer. The scale of adoption is precisely what has alarmed patient advocacy groups, who argue that speed-optimized AI review systems create structural incentives to deny first and let patients or providers absorb the burden of appeal.
Why State Regulators Are Moving Faster Than Congress
While the congressional hearing generated headlines, much of the concrete regulatory action on algorithmic claims denial is happening at the state level, where insurance commissioners have direct authority over utilization management practices within their borders. Several states have already begun drafting or passing legislation requiring insurers to disclose when AI is used in a coverage determination and mandating that a licensed clinician, not an algorithm alone, make the final denial decision for certain categories of care. That state-by-state approach means insurers operating nationally may soon face a patchwork of differing AI disclosure and human-review requirements, adding operational complexity on top of the reputational and legal pressure generated by the federal hearing and pending lawsuits.
The Insurance Industry’s Defense
Insurers generally argue that AI tools are used to flag claims for review or to support, not replace, licensed clinical staff who make final coverage decisions, and that automation actually speeds up approvals for the vast majority of routine, medically uncontroversial claims — freeing human reviewers to focus attention on complex or borderline cases. UnitedHealth has pointed to internal projections that AI-driven efficiency, including in claims processing, will save the company nearly $1 billion in 2026 alone, savings executives argue ultimately help control premium growth across the system.
What Patients and Providers Are Reporting
Physicians describe a pattern where AI-flagged denials arrive with boilerplate justifications disconnected from a specific patient’s clinical picture, forcing already overburdened medical staff into time-consuming appeals processes just to secure care that was, in many cases, ultimately approved on review — evidence, critics argue, that the initial AI-driven denial added friction and delay without adding clinical value. Patient groups say this dynamic disproportionately harms patients with complex, multi-condition needs, who are exactly the population algorithmic systems trained on aggregate claims patterns may handle worst.
Where the Fight Goes From Here
With active litigation, congressional scrutiny, and state legislatures increasingly drafting laws specifically targeting algorithmic claims denial, insurers are under mounting pressure to disclose more about how their AI systems work and to guarantee some baseline of individualized human review. The outcome of the pending court cases against UnitedHealth and Cigna could set a legal precedent for how far insurers can lean on automation in coverage decisions — and how much liability they bear when an algorithm gets it wrong.
