The PCCP Is the Real Decision, Not the Generative AI Docket

The FDA has authorized roughly 1,500 AI-enabled medical devices, and most sponsors treat each clearance as a finish line (Clinical Trial Vanguard). That framing is outdated. For any device with a model that will be retrained, retuned, or updated after launch, the finish line moved in December 2024, when FDA issued its guidance on Predetermined Change Control Plans, or PCCPs, for AI-enabled device software functions. The guidance applies across De Novo, PMA, and 510(k) pathways and requires sponsors to pre-specify how their algorithms will be modified after clearance, rather than filing a new submission every time the model changes (Clinical Trial Vanguard).

This is the operational decision facing device sponsors today, separate from and more urgent than the ongoing debate over generative AI frameworks. A PCCP is not paperwork. It is a binding commitment about the boundaries of future model drift, the validation protocol for each change, and the impact assessment sponsors owe FDA before they ship an update. Get it right at submission, and a sponsor can iterate a diagnostic algorithm for years without returning to the agency. Get it wrong, or skip it, and every meaningful model update becomes a new regulatory event, with the cost and delay that implies. BioSpace’s review of AI-enabled Software as a Medical Device confirms that new authorization routes built around this kind of continuous-modification logic are becoming the norm, not the exception (BioSpace).

The catch is that a PCCP only works if the sponsor already has the internal machinery to execute it: version control tied to clinical validation, documented performance monitoring, defined rollback triggers. That is precisely the governance layer that most AI vendors and health systems are still assembling from scratch. Sheba Medical Center’s arrangement with OpenAI is a candid admission of this gap. As its AI lead put it, “there is some regulation, of course, but we are in a way building the regulation, the guardrails, the monitoring, the policy, the governance… we’re building it together” (Drug Discovery and Development). A hospital-scale deployment site is doing, informally, what a PCCP asks device sponsors to do formally: pre-commit to a change management discipline before the changes happen.

The commercial pressure to move fast is real. GlobalData estimates AI can deliver $21 million in gains per drug development programme, and AI-linked venture financing has grown more than 400 percent since 2014, which explains why sponsors are eager to treat model updates as routine engineering rather than regulatory events (Clinical Trials Arena). A well-built PCCP is what lets a sponsor capture that speed without gambling on enforcement discretion.

For regulated buyers, the near-term test is not whether FDA eventually publishes a generative AI framework. It is whether current submissions include a PCCP that a change management system, ideally documented under something like ISO 42001, can actually execute. Sponsors treating the PCCP as a check-box exercise are building devices that will need a new submission the first time the model meaningfully changes. Sponsors building the governance infrastructure now are buying years of runway. The clearance was never the finish line. The change control plan is.

---

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 PCCPs represent the operationally decisive regulatory moment rather than initial clearance—is coherent and well-supported, though the claim that generative AI frameworks are
Source & Claim Verification	Qwen · local	cleared. Most factual claims are supported by citations, but a few lines lack specific references, such as the statement about the commercial pressure to move fast and the growth in AI-linked venture financing
Regulatory & Framework Fidelity	Mistral	cleared. The briefing accurately reflects FDA’s PCCP requirements for AI-enabled medical devices but underreferences ISO 42001 and EU AI Act alignment, particularly on governance and risk management frameworks
Technical Accuracy	Llama	cleared. The article accurately reflects the FDA’s guidance on Predetermined Change Control Plans (PCCPs) for AI-enabled device software functions and its implications for device sponsors.
Bias, Balance & Hype Control	Gemini	cleared. The briefing effectively counters potential vendor hype by emphasizing the operational challenges and regulatory realities of PCCPs over superficial excitement about AI, though it could benefit from e
Novelty & Non-Duplication	Grok	held. The piece is largely a reframe of the already-published Clinical Trial Vanguard/BioSpace PCCP and SaMD coverage plus the Dec 2024 FDA guidance, not a net-new fact pattern or angle versus the wire.
Validation	DeepSeek	cleared. The central claim that the PCCP is the critical operational decision is logically sound and supported by the new FDA guidance, but its absolute primacy over the generative AI docket is an opinion, not

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