
The U.S. Department of Health and Human Services has published a federal roadmap for clinical artificial intelligence, a document hosted by the Centers for Medicare and Medicaid Services and dated Friday 13 November 2026, according to Nextgov. It proposes state sandboxes, risk-based FDA review, outcomes-based Medicare payment and post-market monitoring, and sets a “better than today” safety bar that clinical AI researchers, academic medical centres and research offices will need to plan around.
What happened
The roadmap is titled “A Clinical Super Intelligence Roadmap: Better Answers, Faster Care, and Lower Costs” and runs eight pages. It uses the term clinical Super Intelligence (SI) throughout and states that SI is the former “AI”; Nextgov, which reported it on 16 November 2026 as a roadmap to boost clinical AI, uses the older term. The PDF is hosted on cms.gov and, in its own words, “does not mandate SI adoption,” “does not create a new agency or duplicate an existing regulator,” and does not authorize autonomy that has not cleared the “better than today” bar for its risk tier.
The document is organised around four priorities: earn public trust, build the data foundation, establish clear rules of the road, and pay for what works. It argues that clinical AI can widen access and lower costs, and it cites early comparisons with physicians, including OpenAI’s o3 model ranking the final diagnosis first in 60% of complex cases against 15.9% for internal medicine physicians, and Microsoft’s MAI-DxO diagnosing 85.5% of challenging cases against roughly 20% for practising physicians. Those figures are the roadmap’s own citations of published studies and company reports, not new evidence generated by HHS.
Key details
The roadmap’s “rules of the road” section lists concrete regulatory directions:
- Low-risk tools: a continued rapid path to market for consumer wellness and clinical decision support software, consistent with FDA guidance the roadmap dates to 2026.
- Competency-based review: FDA regulation aligned to risk, “potentially anchoring around a novel competency-based approach, similar to how physicians are evaluated for their credentials,” for certain SI-enabled devices. Nextgov reports that an FDA discussion draft would require chatbots and generative AI to pass doctor-like competency tests before marketing.
- Post-market monitoring: real-world surveillance and predetermined change control plans so tools can improve while staying accountable for safety, effectiveness and performance, with FDA and CMS collaborating on surveillance infrastructure.
- State sandboxes: partnership with states on SI sandboxes in areas of state-only regulation.
- Liability: clearer regulatory pathways that the roadmap says will lead to clearer ownership of liability, with federal guidelines aiming to offer best practices for de-risking innovation.
The payment section commits CMS to outcomes-based approaches. CMS anticipates extending its ACCESS model to primary care, creating structures for SI-led solutions to join CMS Innovation Center primary care and total-cost-of-care models, exploring outcomes-based payment in Original Medicare, and working with states, including in Medicaid, on alternative payment tests. SI-led solutions would need partnerships with clinicians and facilities for escalation to human clinicians.
Background and context
The roadmap builds on a sequence of federal steps it lists: the White House AI Action Plan of July 2025, OMB memoranda M-25-21 and M-25-22, and an HHS strategy in December 2025 followed by a request for information. On the data side, it says the CMS Health Tech Ecosystem grew from more than 60 initial company commitments to over 800, and that patient digital data access rose from roughly 5% to 60% in a year. The Medicare App Library launched in 2026.
On the regulatory side, the roadmap points to FDA’s TEMPO pilot and a January 2026 clinical decision support guidance. FDA’s December 2025 announcement describes TEMPO as a voluntary pilot from its device centre in which manufacturers may request enforcement discretion for certain requirements, such as premarket authorization, in exchange for collecting and sharing real-world performance data, and it runs alongside the CMS ACCESS model. CMS describes ACCESS as a voluntary 10-year Innovation Center model that pays for technology-supported chronic care based on outcomes; it began on 5 July 2026 and adds heart failure, COPD, substance use disorder and tobacco cessation tracks on 1 April 2027. The roadmap also cites FDA work with ARPA-H on the ADVOCATE programme to develop ways to evaluate highly autonomous systems. FDA has separately adopted the Super Intelligence terminology, covered in our report on FDA’s terminology change.
Reaction
Nextgov reports that the roadmap says the federal government wants to work with states on “areas of clear state regulation” without defining that phrase, and notes that states regulate the practice of medicine and license clinicians. It points to Utah, which is testing through phased pilots whether AI can safely issue initial prescriptions and refills, and reports that the roadmap does not say whether FDA oversees those pilots. Utah’s Office of AI Policy told Nextgov it “remains dedicated to actively coordinating with federal partners.”
What it means for researchers, universities and research offices
Evidence design. The “better than today” standard, staged deployment and tiered autonomy make comparator choice central. A clinical AI study that benchmarks against current care, rather than against perfection, speaks directly to the roadmap’s test; the document also says it “will not wait for flawless evidence to open pathways, but it will require evidence to keep them open.” Teams planning validation studies should expect real-world, post-deployment performance to matter as much as pre-market accuracy, and should read the roadmap’s emphasis on real-world evidence in that light.
Academic medical centres. Health systems running or hosting AI tools may be asked for the monitoring data the roadmap envisions. The roadmap does not specify a reporting mechanism, so any institutional monitoring plan should anticipate the FDA and CMS surveillance infrastructure it describes rather than a fixed template. Where tools reach patients through ACCESS and TEMPO, organisations must be Medicare Part B-enrolled and name a medical director, per CMS.
IRBs and consent. The roadmap names patient control of health data and consent to its use in SI as a core principle. Researchers and an institutional review board reviewing studies of SI-enabled tools should treat consent language, data reuse and escalation to human clinicians as live review items. HHS has also filed a final HIPAA rule changing patient record access, summarised in our HIPAA rule report.
Data sharing and sandboxes. The roadmap’s interoperability and “secure data pathways” commitments favour projects that can use patient-authorised data exchange. Multi-state studies should watch how state sandboxes interact with FDA authority, since the roadmap says some FDA actions may trigger express preemption of differing state requirements.
Payment pathways. Outcome-aligned payment means investigators building economic evidence should measure the outcomes CMS pays for, such as chronic condition control, rather than visit volume.
What happens next
The roadmap describes directions rather than dated rules, so near-term milestones come from the existing programmes. ACCESS accepts applications on a rolling basis and adds new clinical tracks on 1 April 2027. TEMPO’s statements of interest opened in January 2026, with FDA planning to select up to about ten manufacturers in each of four clinical use areas. The competency-based approach is described as potential, the FDA discussion draft Nextgov mentions is a proposal, and the roadmap anticipates CMS creating structures for SI-led accountable care. Research offices should track FDA and CMS Innovation Center announcements for those steps.
Sources
- Centers for Medicare and Medicaid Services: A Clinical Super Intelligence Roadmap: Better Answers, Faster Care, and Lower Costs (PDF), November 2026, checked 2026-11-18
- Nextgov/FCW: Trump administration releases roadmap to boost clinical AI, 16 November 2026, checked 2026-11-18
- CMS: ACCESS (Advancing Chronic Care with Effective, Scalable Solutions) Model, checked 2026-11-18
- U.S. Food and Drug Administration: FDA Launches TEMPO digital health pilot, 5 December 2025, checked 2026-11-18







