The Doctor Test for AI Devices

A different kind of signal

Most coverage of AI medical device regulation has settled into a familiar refrain: FDA hasn’t finished its rules, Europe is stacking new obligations on top of old ones, and manufacturers are stuck waiting. That framing is true but incomplete. The more useful signal isn’t that FDA guidance is unfinished. It’s the shape the agency is reportedly testing while it finishes it. FDA is weighing whether to assess generative AI-enabled devices the way it assesses a physician, through demonstrated competency and ongoing performance review rather than a single clearance event (Axios; MobiHealthNews). A senior FDA digital health official has said publicly that generative AI guidance is coming, which means this framing is not a stray trial balloon (STAT).

That distinction matters more than the timeline. A clearance model built around one-time evidence packages and a competency model built around continuous assessment require different engineering investments, not just different paperwork.

What clears today versus what will be asked tomorrow

J&J’s Monarch robotic system recently won FDA clearance for AI features under the existing device framework (MedTech Dive). That clearance is real and it is useful evidence that today’s predicate-based pathway still works for bounded, well-characterized AI functions. But it is evidence for the system FDA is actively reconsidering, not the one it is moving toward. Manufacturers optimizing purely for a Monarch-style dossier are building for the exam that’s ending, not the one being drafted.

Europe is already living the competency logic

The doctor-style analogy sounds novel for FDA, but Europe’s SaMD regime already operates close to it in substance. Under MDR and IVDR, software manufacturers must demonstrate clinical safety and performance on an ongoing basis while separately aligning with emerging AI-specific requirements, an obligation that stacks rather than resolves into a single static submission (htworld). The two regimes are not sharing a template. But both are converging, independently, on the idea that AI-enabled devices need to keep proving themselves after they’re on the market, not just before.

What continuous assessment looks like without a rulebook

Sheba Medical Center’s experience as OpenAI’s first international hospital partner is the closest thing available to a field test of this logic. Sheba’s AI lead has described building the monitoring and guardrails in parallel with deployment, because no existing guideline reaches the workflows in use (Drug Discovery and Development). That is what competency-style oversight looks like when nobody has written the exam yet: instrumentation built for scrutiny that hasn’t been formally specified.

The decision that matters now

Regulated developers should stop asking whether their next submission clears the current bar and start asking whether their monitoring infrastructure could survive an ongoing competency review, not just a point-in-time file. That means investing in real-world performance telemetry and change-control logic built to be inspected repeatedly, not archived once approval lands. The FDA’s doctor analogy may never become formal guidance in that exact language. But the direction it points, toward devices that must keep demonstrating fitness rather than simply earning it once, is where both major regimes are already leaning.

---

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 argument—that FDA’s emerging ‘doctor-style’ competency model requires different engineering investments than one-time clearance—is coherent and well-supported, though the claim that Europe
Source & Claim Verification	Qwen · local	cleared. Most claims are well-supported with citations, but a few lines lack direct references, such as the discussion on the distinction between clearance and competency models and the implications for manufa
Regulatory & Framework Fidelity	Mistral	cleared. The briefing accurately reflects the evolving regulatory trends in ISO 42001, EU AI Act, and FDA/MDR/IVDR frameworks, particularly the shift toward continuous competency-based assessment, but lacks ex
Technical Accuracy	Llama	cleared. The article accurately conveys the shift in FDA’s regulatory approach towards AI-enabled medical devices, emphasizing continuous assessment and competency review, although some technical details are l
Bias, Balance & Hype Control	Gemini	cleared. The briefing effectively identifies and counters potential vendor hype by distinguishing between current and future regulatory landscapes, emphasizing the shift towards continuous assessment over one-
Novelty & Non-Duplication	Grok	held. Core hook and framing are straight repackaging of the Axios/MobiHealthNews/STAT wire on FDA’s doctor-style genAI review, with Europe/Sheba add-ons that restate known continuous-performance logic rathe
Validation	DeepSeek	cleared. The central claim that FDA is moving toward a ‘doctor-like’ competency model for AI devices is plausible based on reported agency discussions but remains speculative as no formal guidance or rule exis

Sources cited: 15. Validation challenges: 0. Review cost: about $0.04. Learn how these briefings are written and verified.