FDA's AI Perimeter Is Widening and Narrowing at the Same Time

FDA is not moving in one direction on AI. It is moving in two, and the space between them is where regulated buyers now have to make their hardest classification calls.

The perimeter is widening

The finalized January 2026 guidance on wellness versus medical device status broadens what counts as “low risk,” extending that label to non-invasive, non-implanted wearables that measure physiological parameters once reserved for regulated devices, according to MD+DI. Identical hardware can now sit on either side of the device line depending on labeling and intended use. That is a real commercial opening. It is also a trap for any team tempted to market diagnostic-grade inference under a wellness label to avoid the submission burden.

The mechanics inside are tightening

At the same time, FDA’s December 2024 final guidance on Predetermined Change Control Plans governs how AI-enabled device software can be modified post-clearance without triggering a new marketing submission, a mechanism Clinical Trial Vanguard notes still lacks a companion benchmarking standard for model performance claims. FDA has also issued draft guidance on AI device and software evaluation more broadly, a shift Medical Device Network frames as the agency filling gaps left by the administration’s deregulatory posture on AI innovation. Meanwhile, FDA and EMA have jointly published Guiding Principles of Good AI Practice in Drug Development that put explicit accountability on sponsors for AI-derived insights used in regulatory submissions, per Forbes. And FDA’s new rulebook for AI use in clinical trials replaces years of case-by-case answers with something closer to a standard, though implementation remains, in the agency’s own framing, the hard part, according to Clinical Trial Vanguard.

What a narrow clearance actually looks like

ThinkSono’s FDA clearance for AI-powered ultrasound guidance in deep vein thrombosis evaluation is instructive precisely because it is narrow: a specific patient population, a specific exam type, a specific clinical question, per Venous News. That specificity is now the template. FDA has authorized more than 1,450 AI/ML-enabled devices overall, with nearly 300 clearances in 2025 alone, according to Barchart, and the successful path through that volume runs through tightly scoped indications, not broad wellness claims.

The decision

Compliance leaders should not treat the widened wellness lane as a shortcut for anything that performs diagnostic inference. Classify by clinical function, not by hardware form factor, and build PCCP-ready change control architecture before first clearance rather than retrofitting it after a model update forces a new submission. The EU is watching the same tension, with a proposal to simplify MDR and IVDR burden now working through the European Parliament, per Nature. The regulatory perimeter is being redrawn on both sides of the Atlantic. Sponsors who draw their own lines first will spend less time redrawing them later.

---

Board record

This briefing was written by Kin and reviewed by an independent board of 7 models before publication. Ruling: CLEARED.

Seat	Reviewer	Finding
Chair · Editorial Judgment	Claude	cleared. The central thesis—that FDA is simultaneously loosening perimeter definitions while tightening internal controls—is coherent and well-supported by the cited regulatory actions, though the claim that t
Source & Claim Verification	Qwen · local	cleared. Most claims are well-supported by citations, but a few lines lack specific references, such as the discussion on the EU’s proposal to simplify MDR and IVDR burden.
Regulatory & Framework Fidelity	Mistral	cleared. The briefing accurately reflects FDA, MDR/IVDR, and broader AI regulatory trends, but lacks explicit mapping to ISO 42001 or EU AI Act requirements.
Technical Accuracy	Llama	cleared. The article accurately reflects current FDA guidance and trends in AI regulation, but lacks technical depth in some areas, such as specifics on AI model performance benchmarking.
Bias, Balance & Hype Control	Gemini	cleared. The briefing effectively balances the perceived widening and tightening of FDA’s AI perimeter, using specific examples and external sources to counter potential vendor hype about broad ‘commercial ope
Novelty & Non-Duplication	Grok	held. The widening-vs-tightening frame is a light editorial gloss on already-circulating 2024–2026 FDA wellness, PCCP, AI-trial, and clearance items that the cited trade wire has already covered in essentia
Validation	DeepSeek	cleared. The central claim that FDA’s regulatory posture is simultaneously widening for wellness and tightening for medical devices is strongly supported by multiple, specific, and credible sources detailing d

Sources cited: 10. Validation challenges: 0. Review cost: about $0.04. Learn how these briefings are written and verified.