FDA to open a unit of 13 engineers including devs, experts in AI and cloud computing, to keep pace with changes to medical devices powered by machine learning
FDA's Bakul Patel envisions a new regulatory approach to digital health
Context & Ripple Effects
Bakul Patel's 13-engineer unit looks small in isolation, but it is the origin point of the FDA's decade-long institutional build-out around machine learning: the team was hired in 2017 specifically because ML-powered medical devices change after clearance, breaking the agency's one-review-per-product model. Everything since compounds on it — from the agencywide generative AI tool rolled out to scientific reviewers in 2025 (an internal copilot for reviewers) to plans to deploy AI across all FDA centers by mid-2025 (agency-wide AI deployment).
The arc runs through device clearances too: sleep apnea detection for Apple Watch models, Samsung devices, Google's Loss of Pulse Detection, and AirPods Pro 2 cleared as an over-the-counter hearing aid software device all flowed through the regulatory posture Patel's unit was built to develop. By 2026 the endpoint is visible in the administration's push for an FDA fast track aimed squarely at digital health tech like AI chatbots (the digital-health fast track).
First-order effects
- The unit gives the FDA in-house engineering capability — developers plus cloud and AI experts — so reviewers can evaluate software that updates continuously rather than treating each model change as a new submission cycle.
Second-order effects
- Device makers shipping ML features gain a predictable review path, which is what made the wave of consumer-device clearances (Apple, Samsung, Google) administratively possible; competitors building similar health features calibrate their roadmaps to how quickly this unit clears adaptive algorithms.
Third-order effects
- If the pattern holds through the current fast-track proposals and the real-time clinical-trial data-feed pilot, the FDA completes its shift from gatekeeper to continuous monitor — regulation embedded in the data pipeline itself, with the agency's own AI infrastructure (direct data feed to trial data) becoming part of the approval apparatus.
The trend: Medical-device regulation is moving from episodic premarket review toward continuous, AI-mediated oversight, with the FDA staffing itself as a technology organization to keep pace.