Congress is testing how far AI should reach into Medicare care decisions

Congressional Democrats are trying to stop a Medicare pilot that uses artificial intelligence in prior authorization decisions. According to STAT News, resolutions were introduced in the Senate and House after the Government Accountability Officer found that the WISeR program is subject to congressional approval rather than operating merely as a technical administrative experiment.

WISeR is a sensitive kind of digital test: it does not simply ask whether an algorithm can speed up claims handling, but who carries the risk when software, contractors, and a public insurer jointly shape access to care. The pilot is currently running in six states: Arizona, New Jersey, Oklahoma, Ohio, Texas, and Washington. It targets supplies and procedures associated with fraud, but critics argue that this structure can become a system that delays necessary care for older patients.

The core issue is not AI alone. Prior authorization has long been one of the most contested parts of American health care because an administrative filter stands between a clinician’s recommendation and a paid service. Adding AI to that filter raises the stakes. A decision may become faster, but also harder to inspect. When a patient receives a denial, the question is no longer only which official reviewed the request, but which model, under which criteria, and with what level of human oversight contributed to the outcome.

Democrats argue that prior authorization is already aggressive enough without another algorithmic layer. The research brief cites criticism that Americans are tired of tactics that put needed health care out of reach, and that WISeR could become a program that denies care to Medicare patients so companies can profit. Those claims need to be read carefully: the supplied context does not prove the scale of harm, but the political charge goes directly to the design of the program.

One detail matters sharply: contractors are paid under an undisclosed formula that includes the number of denied procedures. If a system that helps decide access to care also operates inside a payment model where denials factor into compensation, trust erodes quickly. That does not mean every denial is wrong. It means the program needs unusually clear safeguards, publicly understandable criteria, and a real appeal path.

The next step now depends on Congress and the response from the Centers for Medicare and Medicaid Services. The official Medicare context sits at Medicare.gov, while CMS innovation models are handled through the CMS Innovation Center. GAO’s role in federal oversight is described at GAO.gov. If the resolutions reach a vote, the debate will no longer be only about one pilot across six states. It will be about the boundary between automation, cost control, and a patient’s right to understand why care was approved or denied.