- AI robustness and accuracy in support of the AI standardisation request
prEN 18229-5 is expected to become a harmonised standard giving presumption of conformity with the robustness requirements of Article 15 of the AI Act. In the medium term it gives providers and testing laboratories one common, auditable way to describe the intended domain of use, test edge and corner cases, and set stability and sensitivity thresholds from the risk acceptability criteria. Anchoring the methods in the ISO/IEC 24029 series keeps the European approach aligned with international practice, and the clear split between robustness (functional correctness) and cybersecurity (prEN 18282) avoids duplicated or contradictory evaluations. In the long term this lowers the cost of compliance for SMEs, supports a European market for independent AI testing and conformity assessment services, and makes robustness evidence comparable across sectors such as biometrics, health and critical infrastructure.
- New standard development
-
Custom funding impact (2nd Open Call 2029)This fellowship supports my contribution to the development of the AI Act standards, especially AI trustworthiness framework – Part 5: Robustness (EN 18229-5), I provide revisions of comments and follow related meetings.
Value of Research
PhD student @LIS-CNRS-UTLN on biometric template protection + AI Testing Laboratory Director @CLR Labs,
Technical contributor in CEN/CLC JTC 21/WG 4 “Foundational and societal aspects” for harmonised AI standards
