Skip to main content
Shaping Europe’s digital future
  • EVENT
  • Publication 25 February 2025

EU AI Office hosts in-person workshop towards defining best practices for systemic risk evaluations

On April 28, members of the Network of Evaluators and the AI Office gathered in Brussels for an in-person workshop to develop evaluation criteria for assessing systemic risk on general-purpose AI (GPAI) models.

Participants to the Network of Evaluators Workshop, at the Institute International de Physique Solvay in Leopold Park the 28th of April 2025

Participants to the Network of Evaluators Workshop, at the Institute International de Physique Solvay in Leopold Park the 28th of April 2025

Over 80 participants from more than 50 organisations were present in this first in-person workshop. Three sessions took place:

  • "Chemical, Biological, Radiological, and Nuclear (CBRN) Risks": Presented by Froggi Jackson (Deloitte) and Sunishchal Dev (RAND), with panel insights from Hannah Herzig (European Commission-HERA) and James Black (Johns Hopkins University)
  • "Cyber Offense": Presented by Dan Lahav (Pattern Labs) and Vas Mavroudis (Alan Turing Institute), with panellists Jaime Raldua (Apart), Malcolm Murray (SaferAI), and Jair Aguirre (RAND)
  • "Loss of Control/Persuasion": Presented by Marius Hobbhahn (Apollo) and Beth Barnes (METR), with contributions from Mario Fritz (CISPA) and Siméon Campos (SaferAI)

A final session gathered reflections and key points from the day, facilitated by Patricia Paskov (RAND), José Hernández-Orallo (VRAIN-UPV), Tom David (PRISM), Max Hasin (METR), and Charlotte Stix (Apollo). 

The speakers in these sessions were selected from a call for expression of interest on AI Office’s Network of Evaluators.