FDA Signals AI Medical Device Framework, But State Laws Are the Real Barrier to Autonomous Care

The FDA is moving toward a regulatory framework for AI-powered medical devices — but the bigger barrier may be state law, according to a September 2026 analysis from Morgan Lewis published on JD Supra.

FDA's August 2026 discussion paper on generative AI-enabled devices broke relative agency silence on the topic. The paper signals intent but does not establish formal policy.

Key FDA signals from the discussion paper:

• Risk is assessed by what a device does, not how autonomously it operates — informing vs. driving clinical action
• FDA is exploring competency-based clearance, postmarket monitoring, and Predetermined Change Control Plans (PCCPs) for model drift
• Foundation model oversight remains unresolved
• More formal guidance is forthcoming, per FDA digital health leadership

Where the real bottleneck is: state medical practice law

FDA authorization does not guarantee deployment. Most state practice-of-medicine statutes were written for human providers and predate AI. A broad definition like Washington's — which covers anyone "offering to diagnose, cure, advise, or prescribe for any human disease... by any means or instrumentality" — may already capture autonomous AI functions, regardless of disclaimers.

A regulatory gap is emerging: FDA could authorize autonomous AI devices before states permit them to operate without a licensed human in the loop. Only Utah and Arizona have launched regulatory sandboxes to ease these restrictions temporarily.

For organizations building or deploying AI tools in healthcare-adjacent contexts, the takeaway is that FDA clearance is just the first gate. State-level legal review and compliance planning will need to run in parallel, not sequentially.

Read the full article on JD Supra