The FDA lets generative AI reach patients before full approval
A new pilot lets generative-AI medical tools go into real clinical use while the regulator learns from real-world data.
Medical regulation usually runs ‘approve first, deploy later’. In September the US FDA started testing a different order for generative AI: deploy under observation, then learn.
What happened
Through a pilot programme called TEMPO, the FDA accepted a first group of generative-AI medical devices, at least four, including tools from Cadence and Limbic. These can reach patients without the usual prior marketing authorisation. The pilot is tied to a Medicare payment model for managing chronic conditions, so these tools can also get paid for.
Why it matters
Generative AI changes too quickly for review cycles that take years. By watching real-world performance instead of static submissions, the regulator is testing a model that other sectors, and other countries, may copy.
What it means for your business
Healthcare providers and medtech companies should get ready for faster, evidence-as-you-go routes to market. That puts a premium on the ability to monitor AI outcomes in live settings: data pipelines, safety signals and clinical oversight built in from day one.
Sources: STAT News (Sep 3)