Senate Democrats move to repeal Medicare's AI prior-authorization pilot

Twenty Senate Democrats have introduced a resolution to end a Trump administration pilot program that uses artificial intelligence to approve or deny medical claims for people on Traditional Medicare. The pilot is called the WISeR Model, short for Wasteful and Inappropriate Service Reduction, and it was launched by the Centers for Medicare & Medicaid Services (CMS) under Administrator Mehmet Oz. The resolution was led by Senate Finance Committee Ranking Member Ron Wyden along with Senators Patty Murray, Maria Cantwell, Richard Blumenthal, and Kirsten Gillibrand, and it has been joined by fifteen additional co-sponsors. A companion resolution has also been introduced in the House by Representatives Greg Landsman and Suzan DelBene.

The push follows a determination last week by the Government Accountability Office (GAO) that WISeR is subject to the Congressional Review Act, a law that lets Congress overturn certain federal agency actions. That determination opens a 60-day window during which Democrats can force a floor vote on a resolution to repeal the model, regardless of whether Republican leadership wants to schedule one. Senator Gallego's office describes the GAO finding as confirmation that the pilot is "currently functioning illegally" across the states where it operates, though the underlying GAO determination itself is described in the sources only as a Congressional Review Act finding, not a ruling on the program's legality more broadly.

Gillibrand, the top Democrat on the Senate Aging Committee, called the use of AI to delay and deny seniors' care "immoral and unacceptable." Wyden said the last thing seniors need is more AI denying the care they need, and framed the vote as a test of whether lawmakers side with seniors seeking timely care or with practices modeled on insurance-industry tactics.

WISeR began on January 1, 2026, and is set to run for six years across six states: Arizona, New Jersey, Oklahoma, Ohio, Texas, and Washington. In each state, Traditional Medicare patients seeking certain procedures must first have their claim reviewed by an AI-driven system before treatment can proceed. If the system denies a claim, the patient must restart the approval process, a delay that Gallego's office says can set patients back weeks. CMS has described participation as voluntary, but Gallego's office says it is effectively mandatory for providers and patients in the selected states because prior authorization is built into how claims are processed there.

Each state's version of the pilot is run by a different for-profit company under contract with CMS. These companies are compensated based on a share of what the program calls "averted expenditures" — money saved by claims that are denied or reduced. Because payment is tied to the size of the denials, the companies have a direct financial incentive to reject more claims, according to the senators' offices, which argue this turns AI-driven medical review into a profit mechanism rather than a neutral check on unnecessary care.

Cantwell's office released a report describing the pilot's impact in Washington state, one of the six states where WISeR operates: since the model was deployed there, patients have waited two to four times longer to complete procedures because of authorization delays, according to Gillibrand's office.

WISeR is not the first place AI has been used to review medical claims. A survey of 93 insurance companies across 16 states, conducted by the National Association of Insurance Commissioners in 2025, found that 84 percent of surveyed insurers use AI across their product lines, and 68 percent reported using or exploring AI for prior authorization approval processes specifically. A 2024 American Medical Association survey found that 61 percent of physicians believed AI use by health plans is increasing prior authorization denials, and large majorities of those physicians reported that denials led to poor clinical outcomes (94 percent) and delayed care (93 percent).

In Medicare Advantage, the private-plan alternative to Traditional Medicare that covers more than half of eligible beneficiaries, a review published in Nature describes similar dynamics playing out at larger scale. KFF research cited in that review found Medicare Advantage plans fully or partially denied 3.2 million prior authorization requests out of 50 million in 2023, and the share of requests denied rose from about 5.6 to 5.8 percent in 2019–2021 to 7.4 percent in 2022. For UnitedHealthcare specifically, the denial rate for post-acute care rose from 10.0 percent in 2020 to 22.7 percent in 2022, a period during which the company was automating more of the review process, according to the same review. A pending class-action lawsuit, Lokken v. UnitedHealth Group, alleges the company used an AI model with a 90 percent error rate to override physicians' treatment recommendations; a federal judge allowed parts of that case to proceed in February 2025.

Some states have already passed laws restricting AI's role in coverage denials. California's Physicians Make Decisions Act, effective January 2025, requires that final medical necessity determinations be made by a licensed physician or health care provider, not by AI or algorithms alone. In 2025, Arizona, Maryland, Nebraska, and Texas each passed similar laws requiring human review, disclosure of AI use, or restrictions on AI's role in denying claims, according to a briefing compiled by the Kansas Legislative Research Department.

At the federal level, CMS rules for Medicare Advantage state that AI-assisted coverage decisions must be based on an individual patient's circumstances and reviewed by a physician or other qualified health professional before a denial is issued. However, when CMS finalized its 2026 Medicare Advantage rule in April 2025, it dropped earlier proposed language that would have required AI-driven services to be provided as equitably as human-delivered ones, saying only that it would "continue to consider" future rulemaking on AI. It remains unclear how much AI regulation or enforcement the current administration will pursue, according to the Nature review, even as litigation against insurers over AI-driven denials continues to expand.