FDA Loosens the Data Bar. The EU Hasn't Said It Will Follow.

Two regulators, two unsettled standards

FDA’s December 2025 update to its real-world evidence guidance lets the agency accept real-world data for device submissions without always demanding a matched control arm, a shift sponsors building AI models on routine EHR and imaging streams are already treating as a green light for retrospective evidence packages instead of prospective trials, per Clinical Trial Vanguard’s analysis. That flexibility is real, and it tracks with the broader momentum behind AI-driven clinical applications, which are outpacing non-clinical use cases in current life sciences market projections.

It is also worth noting that FDA’s own benchmarking posture for medical AI is still catching up to the pace of the field it regulates, a gap Nature Medicine’s 2026 framework explicitly flags. The RWE update is progress, not a finished evidentiary doctrine. Sponsors should read it as directional, not as a settled floor.

What the EU side actually shows, and doesn’t

The claim that EU notified bodies categorically reject FDA-style real-world evidence is not something the current record supports outright. What the record does show is that notified bodies conducting MDR and IVDR conformity assessments, now past 125 dedicated AI reviews at BSI alone, are working from a governance vocabulary built around documented data provenance, bias characterization, and quality management traceability, the same tenets medical imaging AI governance guidance identifies as the operational core of a defensible strategy. The EU AI Act’s Article 10 data governance provisions sit on top of that MDR/IVDR conformity assessment layer, and how the two interact is still being worked out, not finalized, according to Berkley Lifesciences’ compliance mapping. Separately, the clinical literature on AI-assisted decision-making notes that IVDR imposes distinct restrictions on software manufacturers that don’t map cleanly onto a US-first regulatory strategy, per this review of shared decision-making with clinical AI.

Put together, this is not proof of a permanent transatlantic gap. It is evidence of two frameworks moving on different timelines, with the EU side’s rules on data governance still being reconciled across two regulations at once.

The decision this actually forces

For a governance lead, the operative question is not whether FDA’s RWE flexibility is usable. It clearly is. The question is whether to build a training data governance program to the more permissive standard now available, or to the stricter provenance and traceability expectations that MDR/IVDR and Article 10 are converging toward while that convergence is still in progress.

Betting on convergence landing at the permissive end is a bet against the direction every EU signal currently points. Firms that treat data lineage documentation as a first-class deliverable now, rather than a retrofit, avoid re-litigating their evidence package the moment a notified body asks a question FDA no longer requires them to answer.

The unresolved reconciliation between MDR/IVDR and the AI Act is not a footnote. It is the actual sequencing risk on the table, and it is worth pricing in before the first submission goes out the door.

---

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 EU and FDA are on different timelines, not permanently divergent, and that building to the stricter standard now avoids retrofit risk—is coherent and defensible, but the piece u
Source & Claim Verification	Qwen · local	cleared. Most factual claims are supported by citations, but a few lines lack direct references, such as the claim about the unresolved reconciliation between MDR/IVDR and the AI Act.
Regulatory & Framework Fidelity	Mistral	cleared. The briefing accurately reflects FDA and EU regulatory trends but understates the prescriptive requirements of ISO 42001 and the EU AI Act’s risk-based data governance obligations.
Technical Accuracy	Llama	cleared. The article generally demonstrates a good understanding of the regulatory landscape for AI-driven medical devices, but some technical details and citations could be more precisely tied to the claims m
Bias, Balance & Hype Control	Gemini	cleared. The briefing effectively identifies and counters potential vendor hype by distinguishing between regulatory flexibility and settled doctrine, and by highlighting the EU’s stricter data governance requ
Novelty & Non-Duplication	Grok	held. The piece largely restates and cross-links already-published Clinical Trial Vanguard, BSI, Berkley, and Nature Medicine analyses on FDA RWE flexibility versus unresolved EU MDR/IVDR-AI Act data rules,
Validation	DeepSeek	cleared. The briefing’s central claim that the EU has not matched the FDA’s specific RWE flexibility is validated by the provided sources, which show the EU’s framework is still being reconciled and is built o

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