The federal government is testing whether artificial intelligence can help Medicare identify unnecessary medical care, speed up the prior authorization process, and reduce healthcare spending. However, newly released documents are raising questions about whether the program is ready to deliver on those promises. 

The documents, released last week, were obtained by the Electronic Frontier Foundation (EFF) following a March 2026 Freedom of Information Act lawsuit against the Centers for Medicare & Medicaid Services (CMS). EFF is a San Francisco-based nonprofit organization that advocates for digital privacy and civil liberties in the use of technology. 

The new documents detail early problems with the federal government’s Wasteful and Inappropriate Services Reduction (WISeR) model. Launched January 1, 2026, WISeR enlisted six private technology companies to help review prior authorization requests for certain Medicare services. The voluntary model is planned to run in six states – New Jersey, Ohio, Oklahoma, Texas, Arizona, and Washington – until December 31, 2031. According to the CMS webpage, the model is meant to “…protect American taxpayers by leveraging advanced technologies…along with human clinical review to ensure timely and appropriate Medicare payment.” 

According to EFF, the model  requires PA for more than a dozen procedures. They include  steroid injections for pain management, nerve stimulators and treatments, arthroscopic knee surgery, skin substitutes, sleep apnea treatments, and incontinence devices.

EFF’s investigation – and the documents themselves – reveal potential problems with CMS’ AI experiment. 

A rushed rollout from the start
According to EFF’s report, CMS rushed the launch of WISeR before vendors were ready. Innovaccer, one of the six participating vendors, warned CMS that its provider portal might not be fully ready before launch. The records suggest that implementation issues were widespread across vendors, not isolated to one company. 

CMS questioned potentially inaccurate denials
In a March 17 letter to another vendor – Virtix Health – CMS raised concerns about denial rates. Vitrix rejected 53% of the more than 6,000 prior authorization processing requests it received over three months, which the company blamed on issues processing an influx of claims. That figure doesn’t inherently insinuate that the denials were wrong, but it does prompt questions about whether the review process was applying coverage rules accurately, especially if it was processing a large volume of claims.

Authorization delays could affect access to care
CMS flagged vendors for missing turnaround times. One prior authorization request sent to the vendor Genzeon lingered unanswered for 83 days, according to the documents released by EFF.  These delays matter – long processing times may leave patients waiting for treatments to relieve pain or alleviate a worsening condition.

Some provider technology wasn’t ready
The records document CMS’s concern over Innovaccer’s failure to have a fully functioning provider portal for WISeR providers and suppliers. Despite the company’s warning that some of its software might not be ready, CMS maintained its launch timeline. This raises concerns over the readiness of technology being used in a program which affects access to medical care.  

Providers struggled to get answers
In a February letter from CMS to the vendor Zyter, the agency expressed concern that the company may lack adequate processes for responding to provider concerns. If providers can’t get answers about authorization requests, the process becomes even more of an administrative burden.

The payment model raises questions about incentives
WISeR vendors can receive a percentage of the expenditures linked to care that was deemed wasteful and subsequently prevented through their reviews. Though CMS says that compensation is adjusted based on accurate performance, this payment system still creates potential concern for the incentivization of inappropriate denials. 

“Human-in-the-loop” may not be enough
CMS requires a licensed clinician to review every denial recommendation. However, more transparency about how these safeguards operate could help establish whether they adequately protect patients. EFF expressed concerns that the records provide limited insight into whether human reviewers have meaningful ability to change the AI-assisted recommendations.

WISeR addresses a legitimate concern: unnecessary procedures can expose patients to complications while increasing Medicare spending. In incorporating AI, CMS’s hope was that AI technology would be able to help streamline reviews and cut those costs. The early records, however, suggest that the program may have been rushed.