Deployed inside your own EU or EEA environment, not routed through a vendor's cloud outside it. The regulatory environment is the EU AI Act layered on MDR/IVDR conformity assessment, and GDPR's rules on where data is processed — not just where it's stored.
EU AI Act. Most clinical and diagnostic AI lands in the Act's high-risk tier, because it is, or is a safety component of, a medical device subject to third-party conformity assessment. That classification drives most of what a healthcare AI deployment in the EU actually has to do. See the EU AI Act and healthcare AI.
MDR / IVDR. The AI Act's conformity assessment is designed to run through the same notified-body process already used for medical device and in-vitro diagnostic approval, not as a second, separate track — but it adds substantial requirements to that existing process: a lifecycle risk management system, data governance, technical documentation, and human oversight measures. GAMP 5's risk-based model is the common reference point for validating a system that behaves probabilistically. See GxP validation for AI systems.
GDPR. Transfer rules under GDPR cover transient processing, not just storage — sending a document to a model served outside the EEA for inference is a transfer, even if nothing is ever stored there. The same architectural question applies whether the law in question is GDPR or a sectoral data-protection statute elsewhere: does content leave the boundary you chose, for inference, logging, telemetry or monitoring, at any point. See data residency for healthcare AI for how that question is answered in practice.
None of this is legal advice — specific classification, conformity assessment, and data-protection decisions should be made with qualified counsel for your system and jurisdiction. For how the pieces fit together as one governance problem rather than three separate compliance exercises, see healthcare AI governance.
Validation, classification, audit trails, human oversight and hallucination control as one design — covering what EU AI Act high-risk classification requires without treating it as a separate project from MDR/IVDR conformity assessment. See healthcare AI governance.
Extraction, compliance verification and audit trail for submission-heavy workflows, deployed inside your own environment rather than a hosted platform the documents pass through. See regulatory document automation.
De-identified records and imaging screened against protocol criteria, source data resident with the site throughout. See clinical trials.
A two-phase AI pipeline already runs this pattern for a top-5 pharma's intermediate-AMD study: de-identified records and OCT imaging from a leading European academic eye centre, GDPR-compliant by construction, with source data never leaving the site.
Read the case study →No. Eclypse Pte. Ltd. is registered in Singapore. Deployments run inside your own EU/EEA infrastructure — sovereign cloud, on-premise, or air-gapped — so residency and processing locality follow your environment, not Eclypse's registered office.
Yes. Deployments are built so personal data is processed inside the boundary you choose, with no transfer outside it for inference, logging or monitoring — addressed at the architecture level, not just in a data-processing agreement.
No. The AI Act's requirements are designed to integrate with existing MDR/IVDR conformity assessment, not replace it — both apply, and Eclypse deployments produce the documentation that process requires as a standard deliverable.
The same governance approach used across every deployment — documented intended use, a reconstructible audit trail, meaningful human oversight, and layered hallucination mitigation — covers most of what high-risk classification requires.
Our proprietary AI orchestration platform for healthcare: one engine, a registry of reusable task modules and domain agents, and a governed knowledge base — deployed inside your walls and run by your team.