Cassidy proposes five-year Medicare payment test for autonomous clinical AI

Sen. Bill Cassidy has proposed a Medicare payment experiment for AI that performs clinical work without a doctor reviewing every result in real time. His PATIENT AI Act discussion draft would create a five-year test, pairing a route to reimbursement with patient protections—but leaving key payment and responsibility questions unresolved.

Cassidy, the Louisiana Republican who chairs the Senate Health, Education, Labor and Pensions Committee, released the draft on October 29, 2026. It has not been formally introduced or voted on. The payment experiment is part of a larger proposal covering healthcare AI review, privacy and federal agency responsibilities.

From autonomous clinical work to a payment test

The draft would direct the CMS Innovation Center to run the model, beginning no later than 24 months after enactment. Its focus would be applications capable of performing a clinical function autonomously: software doing part of a clinician’s work without a physician interpreting each result at the same time.

Medicare would select participating applications and providers using published criteria. Those criteria would include clinical promise, patient safety, gaps in the evidence and potential benefit to Medicare patients. Selection would therefore be part of the experiment, rather than an automatic payment entitlement for every autonomous tool.

Alongside that test, the package would require a framework for deciding whether Medicare should cover AI-based services. It also calls for payment methods and billing codes. These provisions address both whether an AI service qualifies for coverage and how providers would bill for it.

Important terms remain unwritten. The draft does not define “autonomously,” identify the conditions or clinical tasks the model would cover, or set payment rates. It also does not specify patients’ out-of-pocket costs. Those omissions leave the proposed experiment’s practical scope and financial terms unresolved.

Patient protections, without a liability settlement

The proposed model includes requirements governing what patients are told and how participating providers handle clinical concerns. It would also preserve an existing route for patients to challenge Medicare decisions.

• Patient notice: Participants would need to provide informed notice to Medicare beneficiaries when AI is part of their care.
• Clinical accountability: Participants would need appropriate accountability and escalation arrangements, although the draft does not specify how those arrangements must work.
• Appeal rights: Taking part in the model could not reduce a patient’s existing Medicare appeal rights.

Evaluation would extend beyond spending. The model would measure effects on patient outcomes, safety, access and patient experience as well as cost. That makes clinical results and patients’ encounters with the tools explicit parts of the proposed assessment.

The draft does not settle who is liable when autonomous AI causes harm. Instead, it would have the Department of Health and Human Services engage a nonprofit organization to examine healthcare AI liability and recommend changes. It also says it does not alter liability under federal or state law.

Broader oversight proposals, then a comment deadline

The same draft would reshape parts of medical-software oversight. It proposes a rolling FDA review process for AI software, alongside provisions addressing device review, certification and monitoring after deployment. Certified third parties could review low- to moderate-risk software. The package also calls for privacy guidance on AI use under HIPAA.

Other provisions would support adoption inside healthcare organizations and government. HHS would work with the National Institute of Standards and Technology and sector stakeholders on training materials. Its chief information officer would build datasets for AI-supported healthcare research, while the Comptroller General would examine proposals for transparency and explainability.

Cassidy is seeking input from industry stakeholders, clinicians and policy experts. His office is accepting public comments through December 1, 2026, at patientai@help.senate.gov, giving respondents a chance to weigh in on the proposed payment pathway and its safeguards before the draft advances.