Medicare's AI Program Must Respond Within 72 Hours — the Documents Show One Answer After 83 Days
On January 1, 2026, the American healthcare system Medicare introduced a program in which AI-assisted reviews help decide which treatments require prior authorization. Only after the Electronic Frontier Foundation (EFF) sued the agency CMS for documents did the first measurable results come to light: roughly 1,000 internal pages show that two of the program's six vendors denied or "did not affirm" 5,944 prior authorization requests in the program's first three months — and one internal status report describes a request that sat unanswered for 83 days. That is far beyond the 72 hours CMS says vendors must respond within.
What WISeR Is
WISeR stands for Wasteful and Inappropriate Service Reduction Model, and was launched by the Centers for Medicare & Medicaid Services (CMS) on January 1, 2026 as a six-year program scheduled to run until 2031. It applies to users of Original Medicare — not Medicare Advantage — in six states: Arizona, New Jersey, Ohio, Oklahoma, Texas, and Washington. In total, it could potentially affect around 6.4 million people.
In practice, WISeR means that certain treatments in Original Medicare now require prior authorization, and that this review is carried out by six private vendors selected by CMS, one per state:
- Cohere Health – Texas
- Genzeon – New Jersey
- Humata Health – Oklahoma
- Innovaccer – Ohio
- Virtix Health – Washington
- Zyter – Arizona
CMS emphasizes that the program does not change Medicare's existing coverage rules, and that emergency care, services requiring hospitalization, and treatments where delay could pose a serious risk to patients are exempt.
How the Review Works
The core of the model is that AI systems are used as a support tool in triaging and assessing requests, while formal decision-making authority rests with humans. CMS says AI does not make the final non-payment decision, and that non-payment recommendations must come from properly licensed clinicians using standardized, transparent, and evidence-based procedures. The requirement recurs throughout the documentation: CMS demands that a human clinician with relevant expertise review every single "non-affirmation" — the technology cannot make this call on its own.
In addition, CMS imposes a time requirement: vendors must respond to prior authorization requests within 72 hours. It is against this standard that the released documents must be read.
What the EFF Lawsuit Revealed
EFF filed its FOIA request on January 29, 2026. As of March 24, 2026, CMS had not released any documents and had exceeded both the expedited and the standard deadline — so EFF sued. Through the lawsuit, roughly 1,000 pages of internal documents were released.
Two findings stand out. First, the documents show that two vendors "non-affirmed" — denied or did not confirm — 5,944 prior authorization requests in the program's first three months. Second, one internal status report cited by EFF describes a request that remained unanswered for 83 days. EFF also highlights delays and technical problems in the material.
Important caveats accompany these figures: the sources do not identify which two vendors are behind the 5,944 denials, and they do not state what share of total requests this represents. There is therefore no basis for calculating a denial rate from the available material. And the 83-day case is one documented example, not documented statistics on delays in general.
The Incentives: Up to 20% of the Savings
A central part of the debate over WISeR concerns how vendors are paid. According to Yahoo News, vendors can receive up to 20% of the savings resulting from care the review avoids. That means vendors' revenue is directly tied to how much treatment is denied or reduced — an incentive structure that matters precisely because CMS simultaneously requires human clinical review of every individual denial.
Why These Services Were Chosen
The focus areas are not random. An analysis from KFF cited by The Financial Express shows that the services WISeR covers accounted for $12.3 billion, or 5.3%, of Medicare's Part B spending in 2024. Skin substitutes alone accounted for $10.3 billion — 83% of spending on the selected services.
Cost growth in this area has been steep: KFF found that average spending per service rose from around $2,300 in 2019 to $21,200 in 2024 — an increase of roughly 820%. Around 1.1 million traditional Medicare recipients received at least one WISeR-covered service in 2024.
At the same time, KFF's assessment is that WISeR's financial impact may initially be modest, because the selected services make up a small share of total Part B spending.
Vendor Performance: Virtix Health Under Corrective Action
The early months have also revealed vendor-specific problems. Virtix Health, the vendor in Washington, initially denied more requests than it approved, and CMS placed the company under a corrective action plan because of delays. According to Yahoo News, citing Ars Technica, Virtix itself said its response times had improved and that the plan was ended on August 14. This is the vendor's own statement, not an independently verified fact.
What We Still Don't Know
Several key questions remain unanswered in the available material. We do not know which two vendors are behind the 5,944 denials, or how the denial count relates to the total volume of requests. The sources also give somewhat different renderings of the time standard — Yahoo News refers to "three calendar days," while The Financial Express quotes CMS's "72 hours" — and this discrepancy cannot be resolved from what is available.
Furthermore, there is no primary documentation from CMS in the material on how vendors' AI systems actually assess requests; the accounts of the technology's role rest on CMS statements and EFF's released documents as reported through secondary coverage. With the program in the middle of its first operating year and contracts running through 2031, these questions will be decisive for how accountability for WISeR's results actually plays out.