CMS publishes a federal roadmap for AI in clinical care, with FDA and payment plans
CMS has posted an eight-page Clinical Super Intelligence roadmap that floats a competency-based FDA review for some AI devices, federal preemption of state rules and outcomes-based Medicare payment for AI-led care.
Docket, covering health regulation and AI in care
, 3 min read
5% to 60%Patient digital record access in one year, per CMS
What to know
- CMS posted the Clinical Super Intelligence roadmap on its Health Tech Ecosystem site this week; the document covers HHS, FDA and CMS plans.
- It says FDA may evaluate certain AI devices with a competency-based approach, compared to how physicians are credentialed, and mentions express preemption of differing state rules.
- CMS says it may let AI-led services join Innovation Center primary care models and is exploring outcomes-based payment in Original Medicare.
The Centers for Medicare & Medicaid Services (CMS) has published a federal roadmap for artificial intelligence in clinical care, which the document calls clinical Super Intelligence, or SI. The eight-page document, posted on CMS's Health Tech Ecosystem site and dated by its page as updated November 13, sets out how HHS, the Food and Drug Administration (FDA) and CMS plan to regulate and pay for AI tools that advise patients and clinicians. OrangeSlices AI reported the release on November 13.
The roadmap is a policy direction, not a rule. It changes no regulation, and OrangeSlices notes it announces no new funding or contracts. The document also says what it will not do: it does not mandate AI adoption, does not create a new agency, and does not authorize autonomy for any tool that has not cleared what it calls a better-than-today bar for its risk level.
What it proposes for FDA review
The FDA section is the most concrete. The roadmap says FDA will keep a fast path for lower-risk tools under its general wellness and clinical decision support guidance, both updated in 2026. For higher-risk tools it floats a competency-based premarket evaluation for certain AI-enabled devices, which it compares to how physicians are credentialed, rather than a review of one fixed version of the software.
It ties continued market access to real-world monitoring after a device is on the market, including predetermined change control plans, the FDA mechanism that lets a manufacturer update a cleared algorithm within limits it agreed in advance (see our explainer on PCCPs). FDA and CMS will build on their TEMPO pilot and ACCESS model, launched together in January 2026, to collect real-time performance data, according to the document.
Two items reach beyond device review. The roadmap says some AI devices could gain authorization through regulatory actions that "may trigger express preemption" of state or local requirements that differ from federal law. It also proposes partnering with states on AI sandboxes for areas only states regulate, and says FDA will advise companies on direct-to-patient access paths for bounded, lower-risk diagnostic and therapeutic products.
What it proposes for Medicare payment
On payment, CMS says it wants to reward results rather than visits or billing volume. The roadmap says CMS:
- anticipates extending the ACCESS model to primary care and more chronic conditions;
- sees a chance for AI tools to deliver many components of the Medicare Annual Wellness Visit;
- anticipates letting AI-led services take part in Innovation Center models for primary care and total cost of care, such as LEAD and AHEAD, as long as they partner with clinicians for escalation;
- is exploring outcomes-based payment in Original Medicare, and wants AI-led plan options in the ACA Marketplaces and Medicare Advantage.
The evidence and numbers it cites
CMS says its Health Tech Ecosystem grew from more than 60 company commitments to over 800, and that patient digital access to records rose from about 5% to 60% in one year. To argue that AI can beat today's care, the document cites studies of OpenAI's o3 model, Microsoft's MAI-DxO and Google's AMIE, mostly on hard case sets or simulated visits rather than routine practice. It also lists risks: inaccurate outputs, uneven performance across populations, commercial conflicts of interest in recommendations, and privacy and cybersecurity exposure.
The roadmap arrives as Congress weighs its own approach. Senate HELP Committee chair Bill Cassidy released a discussion draft of the PATIENT AI Act on October 29, with comments due December 1. Earlier this month CMS and ONC also opened electronic prior authorization testing, and CMS launched a digital diabetes prevention collaborative.
Why it matters
The roadmap is the clearest statement yet of how the federal government wants to treat AI that gives medical guidance: lighter, ongoing review at FDA, federal rules that could override state AI laws, and Medicare money for tools that show results. None of it is binding until FDA issues guidance or CMS writes rules, so companies and states will watch for those next steps. For background on the review pathways involved, see our explainer on FDA clearance, approval and De Novo.
This article is general information about how technology works. It is not medical advice. Talk to a clinician about your own health.
Sources
- A Clinical Super Intelligence Roadmap: Better Answers, Faster Care, and Lower Costs, Centers for Medicare & Medicaid Services, November 2026
- Clinical SI Roadmap, Centers for Medicare & Medicaid Services, page updated November 13, 2026
- Senate HELP Committee Chairman Releases Draft Legislation on AI in Healthcare, AACOM, November 2, 2026
- CMS, HHS and FDA Outline Next Steps for Clinical Super Intelligence Adoption in New Roadmap, OrangeSlices AI, November 13, 2026
AI law and policyDigital health and AI in careFDA and health regulation
Written by an AI correspondent under standards set by editor Jacob Knight. Spotted an error? Tell us and we will fix it.