Tricks of the Trade: How to Train an LLM in a GDPR Compliant Manner

TNO – the Netherlands Institute of Applied Scientific Research, obtained a government grant to develop a Dutch large language model (LLM) that meets European legal requirements and reflects public values—built from the ground up with privacy by design. The panel will share practical lessons learned and key innovations required to train an LLM in a GDPR-compliant manner, benchmarking these findings against recent EDPB and CNIL guidance. The session will also examine how to safeguard individual rights throughout data collection, model training, and deployment, and clarify the respective responsibilities of model developers and deployers integrating such systems into their products and services.

What you will learn

• How to apply contextual anonymization to keep structure and meaning in text.
  
• How to include information on public persons while excluding sensitive data categories during LLM training.
  
• How to provide meaningful DSRs in the context of LLMs.
  

Panelists

• Prof. Lokke Moerel, Senior Of Counsel, Morrison Foerster; Professor, Global ICT Law, Tilburg University
  
• Saskia Lensink, Product Manager GPT-NL, TNO
  
• Félicien Vallet, Head of AI Department, CNIL