9 signals this week: 4 Quality, 2 Clinical trials, 3 Regulatory

FDA's own RFI, open for comment until May 29, is the sponsor's chance to shape what "real-time" means before the criteria are finalized in July, after that, the two live pilots become the template every applicant is measured against.

Map your EDC and eClinical architecture against what TRAVERSE and STREAM-SCLC are already doing: continuous data feeds, patient-level traceability, and validation that happens before database lock rather than after. Those two trials are setting the technical bar, not a future guidance document.

Coming from the EU AI Office's own director rather than a law firm's interpretation, this is as close to the source as the guidance gets before formal publication, worth tracking down through official EU AI Office channels to confirm the specific language before building compliance documentation around it.

Don't wait for the guidelines to clear formal adoption. Map every Annex III system in your EU healthcare portfolio, medical imaging, clinical decision support, patient triage, against conformity assessment, transparency, human oversight, and post-market monitoring now, using the Act text as the fallback if Sioli's guidance isn't independently confirmed in time.

Two things are worth tracking separately here: the 10 joint FDA-EMA principles are dated and confirmed (January 14, 2026), while the Q2 2026 finalization date for the underlying FDA guidance is a secondary-source claim that sponsors should verify against FDA's own guidance-tracking page before building a compliance deadline around it.

Build a credibility assessment file now for every AI tool generating evidence used in regulatory submissions, intended use, input/output characterization, risk-based performance assessment, and documentation of influence on the decision, rather than waiting for the Q2 2026 date to be confirmed by FDA directly.

The document's own framing, sustainability tied to "research quality, participant safety and long-term global health outcomes", gives sponsors language to justify climate-risk documentation to ethics committees as a data-integrity issue, not an optional environmental add-on.

For trials running in climate-exposed regions, add extreme weather, infrastructure disruption, and displacement risk to the trial risk management plan alongside the standard operational risks, WHO's document treats these as quality risks now, not force-majeure footnotes.

"One-day" doesn't mean lower stakes: FDA still issues Form 483s off these assessments and investigators retain authority to escalate scope on the spot, so treat the risk-based selection criteria FDA published as the actual inspection-readiness checklist, not a formality.

Build an inspection risk-profile maintenance programme around the criteria FDA actually named, product type, prior inspection outcomes, operational characteristics, rather than assuming a specific AI model's scoring logic that this release doesn't disclose.

Marketing authorisation holders on the 200-plus-substance Union list should start tracking the MSSG's supply-chain vulnerability assessments now, since those assessments, not just the political agreement's headline, will define which production continuity and stockholding obligations actually apply to each product.

Check whether your EU-authorised products sit on that 200-plus-substance Union list, or are likely candidates for it, then map the Act's production continuity, stockholding, and notification obligations against your current supply network before the agreement moves from provisional to final.

A sponsor with pending deficiency responses, historical warning letters, or open label negotiations scattered across separate FDA systems used to have some benefit of the doubt on cross-referencing; a platform built from 40 consolidated sources removes that.

Run a full FDA submission history review across your portfolio now, open deficiency patterns, outstanding commitments, label revision histories, since Elsa sitting on top of 40 consolidated data sources means reviewers can surface that history without you volunteering it.

Programmes leaning on PBPK, PKPD, or PK-bridging evidence should apply now, while EMA is still gathering pilot experience, EMA says it "will capture and consolidate learnings until this procedure is deemed ready for standard implementation," so feedback obtained now is prospective, on a fixed 70-day clock, rather than a retrospective defense of the model after submission.

Build the documentation package EMA actually asks for, the MIDD evidence assessment table plus model analysis plans and reports, before applying, since incomplete documentation is the most direct way to lose the Day 70 timeline to clarification requests.

Any manufacturer whose QMS documentation still reads as legacy Part 820 rather than ISO 13485:2016 should treat Day 100 as the deadline that already passed: FDA confirms it "updates regulatory procedures and program contacts" under 7382.850, meaning investigators showing up now are trained on the new program, not the old one.

Pull your last QMS gap assessment against ISO 13485:2016 and check whether management review, risk management integration, and design-transfer records were actually closed out, not just identified. Investigators are now working from CP 7382.850, not QSIT checklists.