FDA, CMS scramble to avoid regulatory logjam as more states look to test AI in clinical settings

Clinical work means making or carrying out decisions that affect a patient’s care. Examples include writing prescriptions and making treatment decisions. These tasks are closely tied to medical practice, not merely administrative support or general health information.

The important mechanism is autonomy. An AI that helps a clinician review information remains under human supervision. An AI that independently chooses treatment or issues a prescription performs the clinical task itself. The source identifies this distinction as central to the boundary between state medical regulation and federal device regulation.

This boundary matters because states license clinicians and regulate medical practice. If an AI performs clinical tasks without supervision, the product would likely fall under the Food and Drug Administration’s jurisdiction as a medical device. That can create overlapping authority and make adoption harder for companies and health systems.

States and the federal government are involved because clinical AI can perform two kinds of regulated activity at once. It may act like a clinician by making care decisions, while also functioning as a technology product that could qualify as a medical device. Those authorities overlap rather than fitting neatly into separate boxes.

States regulate the practice of medicine and issue licenses to doctors and other clinicians. The federal FDA regulates medical devices. For example, an AI that writes prescriptions without human supervision could resemble both an autonomous clinical practitioner and a regulated medical device.

That overlap is already creating conflict. The Trump administration has tried to reduce regulatory burden for health AI companies, but the division of authority remains unresolved. Companies and hospitals must therefore consider both state medical rules and federal device requirements when developing or adopting clinical AI.

The article directly identifies two regulatory layers. State governments control the practice of medicine and issue licenses to doctors and other clinicians. The federal government, through the Food and Drug Administration, regulates medical devices. A clinical AI product can touch both areas when it performs tasks normally done by a clinician.

The mechanism depends on what the product does and how independently it operates. An AI that supports a clinician may remain a tool used within licensed practice. An AI that writes prescriptions or performs clinical work without supervision would likely be treated as a medical device, bringing it into FDA jurisdiction.

There is also a federal payment layer. The source highlights how Medicare pays for health technology, although it does not name CMS directly. Thus, two layers directly regulate the product, while federal Medicare policy can affect whether providers can pay for and adopt it.

Supervision matters because it helps define whether the AI is assisting a licensed professional or performing medicine itself. A clinician remains responsible for the clinical task when the system operates under human oversight. Without that oversight, the AI may be acting as an independent decision-maker.

Consider prescription writing. An AI could organize information and suggest an option for a clinician to review. A different system could write the prescription itself. The source says that unsupervised performance of such tasks would likely push the product into FDA jurisdiction because the system could be considered a medical device.

The distinction affects the regulatory path and the burden on developers. It also affects hospitals deciding whether to deploy the technology. The article describes efforts to reduce regulatory burden, but the conflict between state authority and federal authority remains unresolved as clinical AI becomes more capable.

Unclear authority creates a practical problem: organizations may not know which rules apply, which approval path to follow, or which regulator will review the product. That uncertainty can increase legal and compliance work. It can also slow development and delay hospitals’ decisions about adopting the technology.

For example, a company building an AI that writes prescriptions without human supervision may need to consider state rules governing medical practice and FDA rules for medical devices. A hospital considering the system must likewise assess whether its clinicians remain involved and whether the product satisfies applicable federal requirements.

The source describes this as a brewing conflict and a key problem for health AI. The Trump administration has sought to ease regulatory burdens and create new payment methods, but overlapping jurisdiction remains. Until responsibilities are clearer, companies and hospitals must navigate both authorities cautiously.

The FDA and CMS address different questions. The FDA decides whether a product is a medical device and regulates that category. CMS, the federal agency administering Medicare, influences whether Medicare will pay for a technology or related service. A product can therefore clear a regulatory hurdle yet still face a payment hurdle.

For example, a hospital may consider an AI tool for clinical care. FDA oversight concerns the product’s regulatory status and safety framework. Medicare payment concerns whether the hospital or clinicians can receive reimbursement for using it. Payment affects the business case for adoption, even when the FDA is the relevant device regulator.

The source mentions both FDA regulation and how Medicare pays for health technology, but it does not name CMS or explain specific payment rules. The broader implication is that health AI adoption depends on regulation and financing, not on device authorization alone.

Medical licensure is the authorization states give doctors and other clinicians to practice. The practice of medicine covers clinical work, including decisions and actions involved in caring for patients. The article states that states regulate this practice and issue licenses to clinicians.

This system matters for AI because a product that performs clinical work could appear to be acting like a licensed professional. An AI that writes prescriptions or makes treatment decisions without human supervision may therefore challenge the state-controlled boundary. A supervised tool, by contrast, supports a clinician who holds the relevant license.

States have traditionally controlled these areas through their authority over medical practice and professional licensing. The federal government has a different role through the FDA, which regulates medical devices. When AI combines clinical functions with device-like technology, these established roles overlap and produce the conflict described in the source.