2026 EDITION

“Oversee me when you should, if you can?” Accountability, and Agentic AI governance in the EU

May 6, 2026 - 09:00 - 10:00

Room H1

Agentic AI systems, capable of autonomous planning, tool use, and goal execution without minimal to no human involvement, are increasingly deployed in real-world environments. For them to realise their promise of efficiency and scalability, we need to explore adequate governance mechanisms. Agentic AI exposes structural weaknesses in existing AI governance approaches in the EU, particularly around accountability, human oversight, and competence requirements.

This interactive workshop uses accountability and human oversight of AI as a concrete governance lens to examine how agentic AI challenges current EU-level regulatory assumptions and emerging implementation practices and to identify the pathways for further development of relevant AI governance.

What is agentic AI, and in which domains it will be/is increasingly being used?
How does the EU AI Act regulate agentic AI uses? Who should be accountable when actions emerge from distributed agentic workflows?
Where can humans intervene meaningfully without introducing automation bias or placebo compliance?
What evidence is needed to demonstrate responsible use and control?

The discussion is explicitly structured around three intersecting topics listed on the Data week conference:

Trustworthy, responsible, and human-centric AI ,
Skills and AI literacy as a prerequisite for effective oversight and innovation,
Policy and standards implementation, translating legal and ethical expectations into technical and organisational controls.