New Documents Intensify Scrutiny of Medicare's AI Prior-Authorization Experiment

The documents concern Medicare's AI prior-authorization experiment, a program that tests whether artificial intelligence can help decide whether certain services are medically necessary before payment. Prior authorization requires patients or providers to obtain approval from a payer, often Medicare, before a treatment, test or drug is covered. The new papers add to an ongoing review of how that experiment is being governed, including what information is used to guide algorithmic decisions and whether safeguards are adequate.

The report is based on newly released documents, but the publicly visible portion does not identify the number of cases affected, the vendor, the model's error rate, or the specific decisions under review. That limits independent assessment. The available record establishes that the documents exist and that they raise oversight questions; it does not show that patients were harmed or that a particular algorithm denied care incorrectly.

Medicare's interest in AI reflects a broader health care push to automate administrative tasks, including prior authorization, which can delay care and burden clinicians. Algorithms can learn patterns from claims and clinical notes, but those patterns may encode historical disparities or outdated practice standards. In a black-box system, clinicians and patients may see a denial without a clear reason, making appeals harder and making bias difficult to audit.

The stakes are access and fairness. Prior authorization affects older adults, people with disabilities and patients seeking specialty care, where small administrative barriers can interrupt treatment. Without transparent documentation, robust validation and meaningful human review, algorithmic decisions can become hard to challenge. The issue matters because a denial that looks routine in data can become a missed diagnosis, delayed therapy or unequal access for beneficiaries.