Medicare's AI claim approvals hit 'alarmingly high' denial rates, lawsuit says

A federal lawsuit is pulling back the curtain on Medicare's experiment with AI-powered claims approvals, and the early read is not good. The pilot program, which uses artificial intelligence to approve care for Medicare beneficiaries, has produced 'alarmingly high' denial rates and delayed decisions, according to the lawsuit seeking information about the initiative, as reported by MarketWatch. The complaint also cites a vendor's warning that the program was not ready for prime time. For a program designed to speed up access to care, the opposite appears to be happening.

Those findings land directly against the promise of AI in health care: faster, fairer, cheaper authorization. Instead, the lawsuit paints a picture of an algorithm underfoot, hanging up treatment decisions and pushing patients and providers into a slower, more bureaucratic appeals maze. For Medicare Advantage insurers piloting the tool, the outcome could mean regulatory and reputational blowback; for physicians and hospitals, the stakes are immediate - denials mean unpaid care, delayed procedures, and more administrative work to fight for coverage.

To understand what is at stake, a little background: Prior authorization has long been the back-office scuffle of American medicine. Insurers typically require approval before they will pay for expensive drugs, elective surgeries, imaging studies, and other high-cost services. Medicare Advantage plans rely on it heavily to control utilization and spending. The manual process is notoriously slow and opaque, which is exactly why AI seemed like such an elegant fix. In theory, a machine can read medical records, reference clinical guidelines, and hand back a coverage decision in seconds, sparing doctors the phone calls and mountains of paperwork.

But a tool that decides too fast, and denies too often, creates a new kind of harm. A delayed or rejected authorization can delay a biopsy, delay a surgery, or force a family to pay out of pocket while they appeal. The lawsuit does not quantify the denial rate, but 'alarmingly high' is the kind of language that moves markets and moves lawsuits. When patients cannot get timely care, the health plan's savings on the back end become someone else's cost on the front end - often the patient's health.

The vendor warning reported in the lawsuit is the quiet part spoken aloud. It suggests that even the company building or operating the AI recognized gaps in readiness that were not resolved before pilot launch. For executives deploying AI in claims, underwriting, or utilization management, this is a cautionary tale: internal warnings are often buried under launch deadlines and investor expectations. The lawsuit now forces that risk assessment into the open, potentially exposing emails, test results, and go-live decisions that boards rarely see.

The broader arc is just as important. AI adoption in health care is accelerating across the board - insurers are using algorithms for prior authorization, providers are using them for clinical documentation, and startups are building 'autonomous coding' and 'automated claims denial' tools. Medicare, as the nation's largest payer, effectively sets the quality bar. If its own pilot is generating alarmingly high denials, you can expect Congress, patient advocates, and plaintiffs' attorneys to train their sights on every health plan using similar technology.

For decision-makers, the strategic takeaway is clear: automation in coverage decisions is a trust business before it is a cost business. The companies that win will be those that deploy AI with guardrails - human review on the highest-stakes cases, clear appeals paths, and continuous monitoring of denial rates by procedure code. The companies that lose will be those that treat AI as a black box cost-cutter. The lawsuit is still in an information-gathering phase, but the discovery process could expose exactly how the algorithm was trained, tested, and launched. That is the moment the court of public opinion will deliver its own verdict.