AI Gatekeepers in Medicare: WISeR Model Draws Fire for Delays and Denials

Investigations and provider reports reveal widespread dysfunction, raising questions about CMS’s decision to hand prior authorization power to private AI vendors.

On January 1, 2026, the Center for Medicare and Medicare Services (CMS) implemented the the Wasteful and Inappropriate Services Reduction (WISeR) model in six states, giving AI companies veto power over care that had previously been guaranteed in traditional Medicare. Unfortunately, there is growing evidence that this model is hurting patients and providers without providing meaningful savings. According to stories compiled by The Washington Post, The Seattle Times, and providers themselves, the WISeR model is interfering with providers’ ability to provide care, leaving seniors in pain and doctors stuck in administrative labyrinths.

According to an investigation by The Washington Post, patients and physicians experienced weeks- and months-long delays, technology “glitches,” and inappropriate denials under the WISeR model in each of the pilot’s six states. While providers can theoretically opt out of the WISeR program, doing so carries the risk that the same AI companies contracted to do the prior authorization will later reject the claim after care is provided during “pre-payment review” and leave the patients on the hook financially - functionally forcing patients and providers into the WISeR model morass.

This leaves patients and their care teams with no option but to wait, even for necessary care. WISeR operators, like Cohere Health in Texas, claim that its technology is “never used to deny care but rather to automate approvals.” But over the same time period, federal and company officials in Texas estimated that only 62% of prior authorization requests sent to the AI companies were getting approved on the physician’s first try.

WATCH NOW

Prior Authorization: Care, Delayed | EP 3

A deep dive into prior authorization’s toll — from doctors to federal policy—featuring Dr. Wendy Dean, Dr. Seth Glickman and Rep. Suzan DelBene (D-WA) on CMS’s new AI-driven WISeR model.

The consequences of these portals not working is weeks-long delays, pain, or even subpar care; in Ohio, doctors considered halting a particular treatment that required multiple steps, because the AI could either improperly deny or delay either step. In Washington, the Seattle Times interviewed patients whose necessary care, including pain management, was delayed for months due to the WISeR model rollout. In one case, a patient was delayed for over two months from getting a pain management treatment that allowed him to walk without pain.

It’s not just patients paying the price. According to The Washington Post’s investigation, providers in all six states have come forward with frustrations about the WISeR model. The Washington Post’s investigation found that the WISeR portal didn’t even work in Ohio until recently, and an executive of the Ohio Urology Group estimated that requests took two weeks to process - four times as long as the 72-hour window the model proposes. In Arizona, physicians reported spending over an hour on the phone trying to get a peer-to-peer conversation to overturn a denial or provide more information, only to have another weeks-long delay before the conversation actually happens. An Arizona provider, speaking to Medscape, described the system as “completely nonfunctional and unsustainable.”

At the same time as these alarming new allegations about the implementation of WISeR, there is growing consensus that the model itself is flawed. An investigation into the costs and uses of the covered services by KFF found that the savings in the WISeR model were largely because biotechnology corporations had inflated the costs of one of these treatments and the government was overpaying. Essentially, this means WISeR is targeting the wrong people; rather than take corporate price gouging seriously, CMS is blaming patients and providers, degrading the quality of the care as retribution. Importantly, CMS addressed the skin substitute overpayment with changes to the payment policy separate from the this model that went into effect the same day as the WISeR model, calling into doubt whether any savings seen on skin substitutes during the pilot period are truly driven by WISeR.

Separately, a lawsuit filed by the Electronic Frontier Foundation (EFF) highlights the shocking lack of public information about the AI companies running the WISeR model. As we warned at HEALTH CARE un-covered last year, the WISeR model allows these AI companies to keep 10-20% of the “averted expenditures” that come from higher denials, essentially financially incentivizing these companies to deny care. The lawsuit filed by EFF follows a Freedom of Information Act request that went unanswered, asking for basic information about the program including agreements with the AI companies, insight into how bias is systematically screened for, and oversight mechanisms, including how and whether the CMS can penalize these vendors. To date, CMS has not provided any of these records to EFF or made them available to the public. Kit Walsh, Director of AI and Access-to-Knowledge Legal Projects at EFF, provided the following quote:

“The public deserves to know how public health benefits are being administered. Using AI to deny or delay coverage raises serious concerns about fairness and due process. Our lawsuit under the Freedom of Information Act aims to provide the public with some of the information needed to assess this program and hold the government accountable.”

The WISeR model is planned to run for six years and provide a blueprint for CMS to subject more treatments to burdensome prior authorization requirements. However, prior authorization, especially with the inclusion of AI, is one of the most universally hated pieces of the U.S. health care system by providers and patients alike. Politicians from both sides of the aisle have criticized the practice and called for an all-out ban. CMMI clarified in an FAQ about the program that participants “who have a high rate of inaccuracy may be terminated from the model,” and the growing body of evidence shows that the bar for removal has been reached in every state of the pilot. However, given the fundamental flaws and misaligned incentives of this program, even removing all six current participants can’t ensure that patients will be protected. CMMI should take action to penalize the current AI companies withholding necessary care and work to end the model altogether.

Zena Wolf is a researcher who previously conducted congressional oversight for Rep. Alexandria Ocasio-Cortez and Sen. Elizabeth Warren. She received her B.A. and M.P.A. from Clark University and has written for publications including The New Republic.