EU AI Act readiness

The EU AI Act introduces obligations for providers and deployers of high-risk AI systems. Behavior monitoring in retail, public space, and elderly-care environments may qualify under Title III depending on intended use. Mindora is engineering the platform to meet those obligations by architecture, not by retrofit.

How we address Title III obligations

• Risk management (Article 9) — lifecycle risk file per product version, including specialist-ensemble model risks.
• Data and data governance (Article 10) — documented training, validation, and OOD testing data; bias checks per vertical.
• Technical documentation (Article 11) — maintained per product, kept aligned with Annex IV.
• Record-keeping (Article 12) — on-device audit logs of detections, accesses, and exports.
• Transparency (Article 13) — deployer documentation and signage templates aligned with AEPD guidance.
• Human oversight (Article 14) — alerts require human review for any consequential action; no automated sanctions.
• Accuracy, robustness, cybersecurity (Article 15)— benchmarked accuracy (e.g. VadCLIP 0.844 OOD AUC), edge-only attack surface, signed firmware, no inbound network exposure.

Three regulatory vectors converging on edge-only

• GDPR — minimization, purpose limitation, and rights become easier to honor when video never leaves the premises.
• EU AI Act — Title III obligations are tractable when the system has documented, bounded behavior and a clean audit trail.
• Schrems II/III risk on US clouds — mitigated by design when the platform does not require transatlantic transfers of personal data.

For the briefing pack with current technical documentation extracts, write to contact@mindoratechnologies.com.