The Pulse

EFF calls for Medicare AI limits after records show rollout problems

The Electronic Frontier Foundation renewed its call for limits on automated Medicare decisions, citing records obtained through a lawsuit. The records describe early problems with an AI-assisted review program, while Medicare says clinicians must review every non-affirmation.

AI.info Team · · fact-checked · against 5 sources ↓ · Published 12 Nov 2026

The Electronic Frontier Foundation is calling for tighter limits on automated decisions in Medicare after records obtained through its lawsuit showed problems in an AI-assisted coverage review program. In its latest newsletter, the digital rights group said the documents revealed delays, denials of care and reports of patient harm.

The organization sued the federal government in March for information about the Wasteful and Inappropriate Service Reduction model, known as WISeR. Its November newsletter revisits the records, and the October post on them says Congress should place greater legal limits on the use of automated decision-making in Medicare.

How the review works

The Medicare review model began on January 1, 2026, and is scheduled to run through December 31, 2031, in New Jersey, Ohio, Oklahoma, Texas, Arizona and Washington. It covers select items and services, while excluding inpatient-only and emergency services and services that would pose a substantial risk to patients if delayed.

The Centers for Medicare & Medicaid Services uses participating technology companies to assess coverage for selected services. The agency says licensed clinicians determine recommendations for non-payment.

Providers can submit a prior authorization request for selected services or have a claim reviewed after the service but before payment. The program does not apply to people with Medicare Advantage. CMS says the model does not change Medicare coverage or payment policy.

Problems in the records

In October, EFF published more than 800 pages of additional records obtained through the lawsuit. The material included applications from 66 companies, including the six selected for the model, as well as contracts and communications between the agency and vendors.

In the program’s first three months, CMS sent requests for information or corrective action plans to at least four of the six vendors, according to EFF. The requests described delays, technical failures and a lack of communication with providers.

CMS asked Virtix about reported problems including turnaround times beyond established guidelines, apparently inaccurate or inconsistent denials and inadequate customer service. The agency also told Zyter it had received reports that determination letters were not being issued on time.

Another vendor, Innovaccer, received two requests for corrective action plans and a notice of non-compliance concerning its provider portal. The company said it had warned CMS months before launch that full portal functionality might not be ready, EFF reported.

Questions about oversight

The group argues that the available records leave the public unable to judge whether clinician review meaningfully oversees the AI systems or whether vendors use appropriate data. It says information about some safeguards was redacted from released records.

CMS says each non-affirmation must be reviewed by a human clinician with relevant expertise and cannot be made solely by technology. The agency also says it will audit participants for consistency with Medicare coverage criteria and adjust payments for inaccurate determinations.

Participating companies receive a percentage of expenditures associated with averted wasteful or inappropriate care resulting from their reviews. CMS says that percentage is adjusted for performance measures, including provider experience.