Generative AI Isn't Breaking Device Rules. It's Avoiding Them.

The mechanism that matters

Compliance officers evaluating an AI-enabled medtech product have learned to ask about the clearance. The sharper question is what happens after it, specifically what mechanism governs the moment the model changes. FDA’s predetermined change control plan, the framework that let J&J’s Monarch bronchoscopy robot add AI tools inside an existing authorization, and the lifecycle monitoring FDA is testing through its TEMPO digital health pilot, works because both cases involve AI/ML functions with bounded, specifiable outputs. Generative AI does not have that boundedness, and FDA is actively soliciting feedback on how to regulate generative AI medical devices because the existing architecture assumes a fixed output space. MobiHealthNews reports the agency is still weighing which regulatory approach even applies to genAI-enabled devices.

Where the capital is actually flowing

That vacuum is not sitting idle. It is shaping where generative AI activity concentrates, and the pattern is worth more attention than the RFI itself. GenScript reported its AI drug discovery business doubled inside a first-half revenue jump of 27.3 percent, and Chai has moved into an antibody discovery collaboration with Bristol Myers Squibb. Neither of these is a clinical device. Both are generative AI operating as a research tool inside drug discovery, a category that sits outside SaMD classification entirely and therefore outside the change control question FDA has not yet answered.

This is not a coincidence of timing. It is a rational allocation of generative AI capability toward the regulatory space that already accommodates it, while clinical deployment, where a PCCP-equivalent does not yet exist, stalls behind the RFI. For diligence teams, that migration is the real signal. A vendor’s decision to position a generative model as a discovery tool rather than a clinical decision support feature may reflect genuine product design, or it may reflect classification chosen to avoid a mechanism that does not exist yet. The two look identical on a slide.

The EU side is not standing still either

The same structural tension is visible in Europe. A Q2 2026 life science law update shows regulators refining AI-enabled product expectations on top of MDR/IVDR obligations, and the resulting dual certification burden is already reshaping cost structures across European healthtech. The reshaping of SaMD routes that BioSpace describes is built for models that drift and update, not for systems with open-ended output.

What this means for the diligence desk

Two screens, not one. For narrow AI/ML features, ask for the live PCCP and its version-level audit trail. For generative AI, ask why the vendor’s product sits where it does on the regulatory map, whether that placement matches actual clinical use, and what happens when FDA finalizes a framework that no longer accommodates the classification the vendor chose. Advisors tracking consolidation in European healthtech are already flagging regulatory readiness as a deal quality differentiator. The same scrutiny belongs on where a vendor decided its generative model would live, not just what it does.

The clearance tells you what a device does today. The classification tells you what the vendor is trying not to be asked.

---

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 core argument—that genAI capital flows toward discovery rather than clinical deployment because regulatory mechanisms for bounded AI don’t accommodate open-ended outputs—is logically sound and wel
Source & Claim Verification	Qwen · local	cleared. All factual claims are traced to citations, but a few sources (e.g., EurekAlert!, Index-Journal, Nurix Therapeutics) are cited without being directly referenced in the text, which could be clarified.
Regulatory & Framework Fidelity	Mistral	cleared. The briefing accurately reflects the current regulatory gaps and distinctions for generative AI under FDA, MDR/IVDR, and EU AI Act frameworks, but does not explicitly address ISO 42001 compliance.
Technical Accuracy	Llama	cleared. The article accurately describes the current regulatory challenges surrounding generative AI in medical devices and the implications for industry stakeholders.
Bias, Balance & Hype Control	Gemini	cleared. The briefing effectively identifies and scrutinizes potential vendor strategies to navigate regulatory ambiguity, directly addressing the core mandate of hype control and counterargument.
Novelty & Non-Duplication	Grok	held. The regulatory-arbitrage thesis (genAI capital migrating into discovery tools to sidestep the missing PCCP-equivalent for unbounded clinical outputs) is a genuine synthesis beyond the cited wire items
Validation	DeepSeek	cleared. The central claim that generative AI is being directed to unregulated research applications to avoid device rules is strongly supported by cited evidence of regulatory uncertainty and capital flow.

Sources cited: 14. Validation challenges: 0. Review cost: about $0.04. Learn how these briefings are written and verified.