Medicare's AI Prior-Authorization Pilot: What 6 Million Patients and Early Records Show
New records describe delays and denials of care and operational chaos in Medicare's test of artificial intelligence for prior authorization.
The Electronic Frontier Foundation published the records on Sept. 8, 2026, focused on problems with the WISeR Model EFF. The disclosure adds operational detail to a pilot that would expand algorithmic review in Original Medicare.
WISeR stands for Wasteful and Inappropriate Service Reduction CMS. CMS, the federal agency that runs Medicare, launched the model to target wasteful and inappropriate services in Original Medicare CMS. The agency said the model will test a new process involving enhanced technologies, including artificial intelligence CMS. The model concerns prior authorization in Original Medicare CMS.
Over 6 million Americans on Medicare will need to get prior authorization from AI for 17 procedures MarketWatch. That requirement sits inside a broader Medicare pilot program testing the use of artificial intelligence to expand prior authorization decisions KFF Health News. Prior authorization is the coverage check before payment. Think of it as a permission check for whether a service meets coverage and medical necessity criteria.
Prior use of AI in coverage decisions has drawn scrutiny. Reported instances of AI-assisted blanket denials of coverage have increased in recent years, particularly for Medicare Advantage plans NIH. A 2024 Senate committee report found that AI tools produce care denial rates up to 16 times higher than typical in some cases MarketWatch. Original Medicare has historically used prior authorization more narrowly than Medicare Advantage. The pilot would narrow that difference.
The broader context here is financial risk from utilization management. Prior authorization is a control on service volume and unit cost. When automated models do first review, direct cost per claim can fall while appeals, rework, doctor friction and payment lag can rise. That tradeoff decides whether cuts to low-value care turn into net savings after admin load and delayed care. For health systems, slow approvals stretch days in accounts receivable and complicate staffing for scheduled procedures.
Looking at what this means for oversight, the WISeR records put validation and auditability at the center. The questions are model inputs, override rates, time to decision, overturn on reconsideration, and documentation burden by procedure code. Problems tend to appear first in resubmissions, peer-to-peer review queues and call-center volume rather than in headline savings estimates. For tracking federal outlays, provider cash flow and vendor performance, authorization cycle time and final denial rates after appeal are more informative than covered lives or procedure counts alone. Contract structure also matters, including who bears the cost of false positives and how performance is measured against no algorithmic review.
