AI Act Regulatory Learning and Standards: Sufficient to Protect Fundamental Rights?

With the political agreement on the AI Act now in place, the spotlight shifts to it technical implementation. Technical standards need to be quickly established, harmonised and adopted by AI developers and the certification and market surveillance authorities that will oversee them. The Act requires harmonised standards be put in place, and European standards bodies such as CEN/CENELEC JTC21 are examining systems of emerging AI standards from bodies such ISO/IEC JTC1 SC42. This must be done in a way that can accommodate the accelerating change in AI technologies and applications and the wide variety of learnings that may come from regulatory experimentation through sandboxes and field trials in different high risk areas. This panel will begin to explore how the complex network of standards and regulatory bodies can cooperate with stakeholders to build a reliable regulated market for AI.

• Which forms of regulatory learning, such as sandboxes, will be the most useful in understanding how technical standards can effectively protect fundamental rights?
• How will horizontal technical standards harmonised to the AI Act be adapted to the varying health, safety and fundamental rights protection of the different vertical high risk areas?
• How can stakeholder-led regulatory learning from AI act sandboxes and user trials be fed back effectively into the standardisation revision process where stakeholders are often absent?
• Will the regulatory learning on fundamental rights protections from different domains and states risk fragmenting the consensus underpinning standards and disrupting the free movement of certified product across and into the single market?