The short version
The Centers for Medicare & Medicaid Services (CMS) posted an eight-page document titled "A Clinical Super Intelligence Roadmap: Better Answers, Faster Care, and Lower Costs" in November 2026. It sets a direction for artificial intelligence (AI) in health care, which the roadmap calls super intelligence (SI). It is not a rule. It creates no billing codes, no coverage changes, and no deadlines for your practice. It never mentions the Health Insurance Portability and Accountability Act (HIPAA). What it does say is that SI "relies on sensitive health information" and requires "the enforcement of gold-standard frameworks for privacy, cybersecurity, data stewardship, and patient consent." That work lands on the practice and its vendors, and HIPAA already covers most of it. (Source: CMS, A Clinical Super Intelligence Roadmap (PDF))
Why a practice owner should read this one
One line in this roadmap stands out for anyone who runs a practice: "Patients should control their health data, understand and consent to its use in SI, and remain empowered to steer their own health care decisions." (Source: CMS roadmap, page 5)
CMS wants AI to move fast. Patients get more control. Somebody has to make that control real, and that somebody is you and the vendors you sign with.
I ask practices a simple question: which of your tools use AI? The honest answer is usually "more than we thought." The scribe. The phone service. The chat window someone used to draft an appeal. This roadmap is a good reason to finish that list.
What CMS actually said
The roadmap is a direction, not a regulation. Its "What Will Not Be Done" section says: "This roadmap does not mandate SI adoption." It also says the roadmap "does not create a new agency or duplicate an existing regulator." (Source: CMS roadmap, page 6)
A few facts about the document:
- It uses new words for an old idea. The roadmap writes "SI" where earlier federal documents wrote "AI." It refers to the White House "SI (formerly AI) Action Plan" from July 2025 and says HHS, the U.S. Department of Health and Human Services, "released its SI (formerly AI) Strategy in December 2025." (Source: CMS roadmap, page 5)
- It sets no dates for you. Its plans are goals: CMS "anticipates," "aims," and "is also exploring." There is no deadline, billing code, or coverage rule.
- It does not mention HIPAA. I searched the full text. The words "HIPAA" and "business associate" do not appear anywhere in it.
Your HIPAA duties are the same as they were before November. What changed is how many places patient data is about to travel.
The three shifts behind the roadmap
1. Patients control their data and consent to its use
Under its Earn Public Trust priority, the roadmap commits to "Protect patient choice, consent, and control over health decisions." (Source: CMS roadmap, page 2)
The roadmap says patient matching, identity verification, and "mandatory provider response requirements" drove patient digital data access "from roughly 5% to 60% in one year." It also describes the Medicare App Library, launched in 2026, where patients can use SI tools based on "the portions of their health records they choose to have considered." (Source: CMS roadmap, page 5)
The lesson for a practice: HIPAA does not require patient consent for most treatment, payment, and health care operations. The Privacy Rule says a covered entity "may obtain consent" for those uses, which makes it optional. (Source: 45 CFR 164.506(b)) Uses outside what the rule permits need a signed authorization. (Source: 45 CFR 164.508(a)(1)) And a business associate "may use or disclose protected health information only as permitted or required by its business associate contract." (Source: 45 CFR 164.502(a)(3))
Your practical question: do you know what each AI vendor is allowed to do with your patients' information, and does it match what your patients were told?
2. National standards instead of a patchwork of state rules
The roadmap says it will "Pursue national standards so innovators and patients are not stuck with a confusing patchwork of state rules." (Source: CMS roadmap, page 2)
That is a goal. The same document says the federal government will "partner with states on SI 'sandboxes' to address areas of state-only regulation." (Source: CMS roadmap, page 7) Where it mentions overriding state law, it refers to the preemption that can apply to certain medical devices authorized by the Food and Drug Administration (FDA).
State rules apply today. Three examples:
- Texas. A Texas law in effect since September 1, 2025 allows a health care practitioner to use AI for diagnostic purposes if, among other conditions, the practitioner "reviews all records created with artificial intelligence," and it requires the practitioner to "disclose the practitioner's use of that technology to the practitioner's patients." (Source: Texas S.B. 1188, Health and Safety Code section 183.005)
- Texas, again. Since January 1, 2026, a separate Texas law requires a provider that uses an AI system "in relation to health care service or treatment" to disclose that to the patient no later than the date the service is first provided. That language is broad enough that it may reach AI scribes, not only diagnosis. (Source: Texas H.B. 149, Business and Commerce Code section 552.051)
- Minnesota. Minnesota's health records law generally bars a provider from releasing a patient's health records without "a signed and dated consent from the patient," unless Minnesota law specifically allows it. That is stricter than HIPAA. (Source: Minnesota Statutes section 144.293, subdivision 2)
The patchwork is still yours to manage. See our running list of federal and state AI rules in healthcare.
3. Paying for outcomes, not activity
CMS wants to reward tools that improve outcomes and lower costs, "not those that simply generate more visits, clicks, or billing." Its example is the ACCESS Model, announced in December 2025, which "rewards clinical results rather than activities." (Source: CMS roadmap, pages 6 and 7)
Paying for outcomes means measuring them, with systems "so performance can be tracked after tools are deployed." (Source: CMS roadmap, page 7) More data will flow to more vendors, and each flow belongs in your HIPAA program.
A five-question check of your AI tools
You do not need a new program. Answer these five questions for each AI tool.
1. Which tools touch patient data?
Start with an inventory. The HIPAA Security Rule requires "an accurate and thorough assessment of the potential risks and vulnerabilities to the confidentiality, integrity, and availability of electronic protected health information." (Source: 45 CFR 164.308(a)(1)(ii)(A)) An AI tool that sees patient information is in scope, including add-ons inside software you already pay for and connectors that pass data to other companies. See a real HIPAA risk analysis for the inventory step and AI chats and connectors for the tools people forget.
2. Is there a Business Associate Agreement with each vendor that handles patient data?
A vendor that creates, receives, maintains, or transmits protected health information (PHI) for you is a business associate. You may let it do that "only if the covered entity obtains satisfactory assurances" that it will safeguard the information. (Source: 45 CFR 164.308(b)(1)) Those assurances go in a written Business Associate Agreement (BAA) that meets 45 CFR 164.504(e) and, for electronic information, 45 CFR 164.314(a). (Sources: 45 CFR 164.504 and 45 CFR 164.314) One agreement can cover several tools from the same vendor if its scope says so, and the vendor is responsible for getting the same assurances from its own subcontractors. (Source: 45 CFR 164.502(e)) A BAA on one product tier does not always cover every feature or connector. Start with who needs a BAA, then ask the question in where is your data.
3. Can you show what patients were told and what they agreed to?
For each tool, write down what patient data it uses and why. If a vendor wants to use your patients' protected health information for its own purposes, such as improving its product, two things must be true: your BAA has to allow that use, and the Privacy Rule has to permit it. A BAA cannot give a vendor more room than the rule gives your practice, apart from narrow exceptions for the vendor's own management and administration and for data aggregation. (Source: 45 CFR 164.504(e)(2)(i)) If either one is missing, the vendor cannot use the information that way unless each patient signs a valid authorization. Data that has been properly de-identified is no longer protected health information, but a vendor may de-identify your data only if your BAA allows it. (Sources: 45 CFR 164.514 and HHS de-identification guidance) Then check your state rules, like the Texas disclosure and Minnesota consent laws above. If you cannot show it, treat it as not done.
4. Who reviews the AI output?
The roadmap builds trust through "appropriate human oversight." (Source: CMS roadmap, page 2) HIPAA does not say who must check an AI scribe's note, though Texas does for diagnostic use. A separate federal civil rights rule, under Section 1557 of the Affordable Care Act, applies to health programs that receive federal financial assistance from HHS, such as Medicaid payments. Its "covered entity" is not the same as HIPAA's. Those programs must make "reasonable efforts" to identify and reduce the risk of discrimination from patient care decision support tools, a duty in effect since May 1, 2025. (Sources: 45 CFR 92.210 and 45 CFR 92.1) On the HIPAA side, the Security Rule requires you to "regularly review records of information system activity, such as audit logs." (Source: 45 CFR 164.308(a)(1)(ii)(D)) Name a person for each tool. Write down what they check and how often.
5. What happens to the data when the contract ends?
Almost nobody asks this at signing. A BAA must require the vendor, at termination and "if feasible," to "return or destroy all protected health information" it still holds and "retain no copies." If that is not feasible, the vendor must keep protecting the information and limit its use. (Source: 45 CFR 164.504(e)(2)(ii)(J)) Ask every AI vendor where your recordings, transcripts, chat logs, and backups live, and how it will prove they are gone. The audit protocol used by the HHS Office for Civil Rights (OCR), last updated in July 2018, does not mention AI, so you have to bring these questions yourself. (See the HIPAA audit protocol and AI.)
What I would do this quarter
Make the list of AI tools. Match each one to its vendor's BAA. Add the list to your risk analysis. Name a reviewer for each tool. Then put the termination question to your top three vendors in writing. None of that requires waiting on CMS.
Want a second set of eyes on your AI tools?
That is a conversation we have every week. We will go through the AI tools your practice uses, look at the vendor agreements behind them, and show you where the gaps are.
Live Compliance has been doing this since 2010, with a 100 percent audit success rate. Audit-ready in 60 days, or we keep working at no additional cost until you are.
Request a 15-minute AI tool privacy check
Prefer to start on your own? The free gap scan gives you an audit-readiness score and needs no signup.
Accuracy & legal note. This article is a plain-language summary of the CMS document "A Clinical Super Intelligence Roadmap: Better Answers, Faster Care, and Lower Costs" as posted at cms.gov, the HIPAA Privacy and Security Rule provisions cited above (45 CFR 164.308, 164.314, 164.502, 164.504, 164.506, 164.508, and 164.514), 45 CFR 92.1 and 92.210, Texas Health and Safety Code section 183.005, Texas Business and Commerce Code section 552.051, and Minnesota Statutes section 144.293, as of November 16, 2026. The roadmap is a statement of direction. It does not change HIPAA and creates no new obligation for providers on its own. Regulations, OCR guidance, and enforcement priorities change. This is general educational information, not legal advice, and it does not guarantee the result of any OCR audit or investigation. Verify current requirements at hhs.gov/hipaa or with your compliance counsel before acting. We are committed to providing the highest level of compliance consulting at all times. We do not practice law, we do not provide legal advice during our instruction, nor should you rely on Live Compliance as legal representation. Last updated: November 16, 2026.
Questions
Frequently asked questions
It is an eight-page CMS document, posted in November 2026, on speeding up the safe use of AI, which it calls super intelligence, in health care. It has four priorities: earn public trust, build the data foundation, set clear rules of the road, and pay for what works.
No. The roadmap does not mention HIPAA or business associates, and it says it "does not mandate SI adoption." Your HIPAA duties are unchanged. Its call for "gold-standard frameworks for privacy, cybersecurity, data stewardship, and patient consent" is a direction, and HIPAA already covers much of that ground.
In this roadmap, yes. The roadmap refers to the White House "SI (formerly AI) Action Plan" and the HHS "SI (formerly AI) Strategy."
Yes, if the vendor creates, receives, maintains, or transmits protected health information for your practice. The contract must meet 45 CFR 164.504(e) and, for electronic information, 164.314(a).
Not today. National standards are a stated goal. State laws, such as the Texas law on AI in diagnosis and the Minnesota health records consent law, still apply.

Jim Johnson
Founder and president, Live Compliance
Jim has run HIPAA programs for 500+ healthcare organizations since 2010. He writes about the compliance work practices actually face: the Privacy and Security Rules, OSHA, and what an auditor asks for.
About the founderNew pieces by email
The next piece, when it is published, and nothing else.
Unsubscribe any time. Nothing here needs an email to read.