WHAT WE DO
The AI Act, adopted by the EU on June 13, 2024, introduces a phased regulatory framework for AI. Implementation begins in February 2025 for prohibited use cases and extends to August 2027 for high-risk AI in regulated sectors. Key compliance pillars include AI regulatory sandboxes, conformity assessment through self-assessment or notified bodies, and market surveillance. Union AI Testing Facilities will support market surveillance with testing, coordination, and standardisation efforts.
NoLeFa-84 is a two-year pilot project starting in late 2024, tasked with laying the groundwork for these facilities. Coordinated by Inria, it involves public bodies, SMEs, startups, and an EU research network. Its activities include:
- Standardisation & Compliance: Contributing to developing harmonised standards at CEN-CENELEC and ISO/IEC by 2025, supporting high-risk AI compliance.
- R&D and Testing Tools: Developing testing tools for AI systems’ performance, but also criteria like bias, robustness, and data quality, with quarterly GitHub updates starting January 2025.
- Authority Support & Training: Building collaboration networks, hosting workshops, and providing technical advice and training to market surveillance authorities.