Roy Betser is a PhD candidate in the Technion and a senior AI researcher in Fujitsu Research of Europe, where he is part of the GenAI Trust team. His research focuses on analyzing representation spaces in foundation models and on developing practical trust and safety mechanisms for agentic AI systems. He works on studying the geometry and statistics of learned latent spaces, and developing efficient and practical methods based on this understanding. Additionally, in the area of agentic AI systems, he works on designing lightweight, deployable techniques for policy compliance and risk mitigation, together with general threat modeling and vulnerability evaluation. His work aims to bridge theoretical insights about model behavior with robust engineering solutions for real-world AI deployment.