Clearance Was Never the Finish Line

Clearance Was Never the Finish Line

Life sciences compliance teams have spent two years optimizing for the wrong milestone. Getting an AI-enabled device through FDA authorization is treated as the hard part. It is not. The hard part starts the day after, when the model keeps learning and the paperwork does not.

FDA’s Predetermined Change Control Plan pathway is the clearest signal yet that the agency agrees. PCCP requires sponsors to pre-specify exactly how their algorithms will be modified after authorization, rather than treating every retraining event as an open question for a future submission (Clinical Trial Vanguard). That is a meaningful fix for a real problem: a model retrained on new data without regulatory review is, functionally, a different device wearing the old clearance (Clinical Trial Vanguard). But PCCP only covers what happens after authorization. It does nothing for the roughly 1,500 AI devices already on the market whose post-deployment behavior most sponsors are not yet instrumented to track.

This is where the compliance calculus gets genuinely hard for regulated buyers, not just sponsors. Hospitals and health systems procuring AI-enabled SaMD are inheriting a lifecycle governance obligation that most procurement contracts do not currently price in. Regulatory routes are diverging by market, and evidence requirements are increasingly framed around what happens to a model’s performance over time, not just its performance at the moment of clearance (Medical Buyer). Buyers evaluating a vendor’s regulatory file need to ask a different question than “was this cleared.” They need to ask “what is the pre-specified boundary of change, and who is monitoring for drift outside it.”

Europe adds a second layer. The Digital Omnibus agreements have realigned EU AI Act high-risk compliance timelines with MDR and IVDR, giving embedded AI in medical devices breathing room through August 2028 (healthcare.digital). That reprieve is procedural, not substantive. Sponsors still face a dual-certification structure where SaMD must satisfy both device-specific rules and AI Act obligations layered on top (healthcare.digital). Industry forecasts already treat 2027 as the year every European clinical AI company spends building quality management systems capable of satisfying both regimes at once (Nelson Advisors, via healthcare.digital).

The decision this creates for regulated buyers is not whether to adopt AI-enabled devices. Adoption is already underway across specialties, including areas like retina screening where AI is expanding trial design and access faster than oversight structures are maturing (Ophthalmology Times). The decision is whether procurement, clinical engineering, and quality functions have a shared mechanism for tracking a device’s permitted drift boundary against its actual post-market behavior. ISO 42001-aligned lifecycle controls are the natural home for that mechanism, but only if they are built before the next model update ships, not after an inspector asks for the change log.

The regulators are building the guardrail for what changes after clearance. The buyers still need to build the fence for knowing when it moves.

---

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. Core argument that post-clearance lifecycle governance is the harder problem is coherent and well-supported, but the leap to ISO 42001 as ‘the natural home’ for buyer-side drift monitoring is asserted
Source & Claim Verification	Qwen · local	cleared. All factual claims are supported by citations, but some lines could be more clearly linked to specific sources for maximum clarity.
Regulatory & Framework Fidelity	Mistral	cleared. The briefing accurately reflects key regulatory expectations (PCCP, EU AI Act alignment with MDR/IVDR, and ISO 42001 lifecycle controls) but omits explicit mapping to FDA’s SaMD pre/post-market balanc
Technical Accuracy	Llama	cleared. The article accurately captures the regulatory challenges and evolving landscape for AI-enabled medical devices, but could be improved with more technical depth on AI model drift and lifecycle governa
Bias, Balance & Hype Control	Gemini	cleared. The briefing effectively identifies and counters potential vendor hype by focusing on post-clearance challenges and regulatory complexities, rather than solely celebrating initial authorizations.
Novelty & Non-Duplication	Grok	held. Core thesis (clearance ≠ finish line; PCCP; post-market drift; ~1,500 uncleared-for-change AI devices; EU dual-track) is already saturated on the trade wire the draft itself cites, and the buyer/ISO 4
Validation	DeepSeek	cleared. The central claim that post-clearance model lifecycle management is the harder, unaddressed challenge is strongly supported by regulatory developments like PCCP and market evidence of existing deploye

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