FDA says formal rules for AI medical devices are coming — but no date yet
A top FDA digital health official confirms the agency is drafting guidance on generative AI in medical devices, covering both broad principles and narrow, high-stakes uses.
What happened: Rick Abramson, who directs the FDA's Digital Health Center of Excellence, told STAT the agency is working on formal policy guidance for medical devices built on generative AI — the kind of technology that can write, summarize, or generate responses rather than just flag a pattern in an X-ray. He said the FDA wants to give the industry the clarity it has been asking for, and is weighing what's described as a 'competency-based approach' to evaluating these tools.
Why it matters: Generative AI is already showing up in tools that draft clinical notes, summarize patient records, or power mental health chatbots, often without a clear FDA pathway telling companies what's required to prove safety. Without written guidance, developers, hospitals, and doctors have had to guess at what regulators will accept, which can slow down legitimate tools or let riskier ones through with less scrutiny.
How it works, plainly: Abramson indicated the FDA plans two layers of guidance: one broad document covering generative AI in medical devices generally, and separate, more specific guidance for particular uses that are unusually complex or high-interest — think tools that make diagnostic suggestions versus ones that just organize paperwork. A 'competency-based' approach would judge tools on what they're actually shown to do well, rather than applying one blanket rulebook to every AI product.
The rollout: No timeline was given for when any of these documents will be published, and the FDA's public comments so far describe an intention rather than a finished plan. That leaves an open window where device makers, clinicians, and patients are relying on the agency's existing, older device rules to cover a technology those rules weren't written for.