Implementing the AI Act: How to Assess Generative Artificial Intelligence?

Since the launch of ChatGPT in November 2022, generative AI has taken the world by storm. Its advance is now set to unfold in a regulated manner following the approval of the AI Act by the European Parliament in March. Expected to be enacted within weeks, the AI Act marks a significant milestone as the continent’s first-ever legal framework on AI.
 
However, many questions remain concerning its implementation. Member States and EU bodies such as the new AI Office will have to fill this void. A critical, yet unresolved issue is how to assess systems using generative AI – or “general-purpose AI” in AI Act terminology. Key questions include: How can we reliably assess the weaknesses and dangers of these systems? When should we trust such assessments? And how can we adapt them as technology evolves? Good answers will not only need good tech, but also good economics. The central challenge is to figure out how to design incentive and coordination schemes in order to achieve reliability, trust and adaptability.
 
Our lunch debate will kick off by proposing a solution to these questions based on the latest research results. Our panel wil then discuss the proposal, the institutional setup required to ensure good governance of generative AI, and its real-world implications.

Registration upon personal invitation only. If you are interested in joining the Lunch Debate, please contact our event team under lunchdebates@zew.de.

This event is supported by the Baden-Württemberg Foundation (website in German language).