TEMPO—a voluntary, pilot regulatory framework—provisionally enables certain manufacturers to launch digital health devices for clinical use before obtaining traditional FDA marketing authorization, while the agency evaluates their performance in real-world settings. The program is a key new element of the agency’s Home as a Health Care Hub initiative, which launched in 2024.

TEMPO is designed around an important premise: some digital health technologies may be difficult to evaluate entirely through traditional premarket pathways because their performance depends on how they are used in real-world clinical environments.

For these technologies, a manufacturer may request that the FDA exercises discretion on the enforcement of certain regulatory requirements when its device is intended to provide care covered by the Advancing Chronic Care With Effective, Scalable Solutions (ACCESS) model—a Centers for Medicare and Medicaid Services (CMS) payment model that expands technology-enabled care for patients with chronic conditions. Rather than broadly exempting devices from FDA oversight, TEMPO utilizes a risk-based enforcement approach where, depending on potential patient impact, certain requirements might be waived or deferred.

TEMPO is currently limited to four clinical areas—early cardio-kidney-metabolic, cardio-kidney-metabolic, musculoskeletal, and behavioral health—and is not intended for speculative technologies without a functional prototype. Participating manufacturers are expected to eventually seek appropriate authorization, potentially using data generated during the TEMPO pilot alongside additional evidence.

Regulatory frameworks have generally struggled to keep up with the rapid pace of generative AI innovation. The traditional evaluation of medical devices assumes that the product’s performance and characteristics are sufficiently characterized before widespread clinical deployment. However, this is often not the case with AI-enabled devices, whose performance may vary dynamically across patients and clinical contexts in ways that are difficult to characterize before deployment.

This makes TEMPO a regulatory learning laboratory as much as a pathway for manufacturers. The FDA can observe how AI systems perform outside of controlled environments. For manufacturers, the attraction is also substantial. TEMPO may derisk the deployment pipeline of new AI technologies by enabling companies to generate real-world evidence while working toward eventual FDA marketing authorization.

In an August 2026 discussion paper, the FDA described a potential “competency-based approach” to evaluating AI products, which would assess a device’s capabilities through benchmarking and clinical confirmation, rather than attempting to evaluate every possible use case before deployment. TEMPO could serve as an early sandbox for the FDA’s evolving approach to AI regulation, with early real-world deployment and competency-based evaluation informing one another.

TEMPO participants are entering a period of real-world deployment and evidence generation. The FDA plans to expand the pilot to include up to approximately 10 manufacturers in each of the four clinical areas—roughly 40 participants in total. The agency has indicated that it hopes this cohort will include a broad representation of manufacturers across different sizes, types, and stages of maturity.

As the pilot expands, manufacturers will collect, monitor, analyze, and report real-world performance data, including data relevant to predefined patient outcomes. The FDA may use these data to better understand the benefit-risk profiles of participating devices and to support manufacturers’ eventual marketing submissions. The agency plans to engage participants through time-limited “sprint discussions” focused on specific regulatory or submission-related questions, potentially creating more iterative interactions between manufacturers and regulators.

TEMPO is not a permanent alternative to FDA authorization; rather, the pilot creates a temporary bridge between development and authorization. The more consequential question is what the FDA learns along the way during this pilot. If real-world evidence can reliably identify safety risks while establishing clinical benefit, then TEMPO could provide a streamlined model for evaluating certain AI-enabled technologies. Conversely, if important problems and perils emerge only after widespread clinical use, this suggests the limits of relying on postmarket evidence.

TEMPO represents an important experiment on how the FDA may regulate AI-enabled medical devices, providing regulators an opportunity to learn from technologies in real time. The lessons from TEMPO may shape not only the future of digital health innovation but also broader FDA frameworks in the age of generative AI.