How to make AI Safe?

​How to make AI Safe?

​📍 University of Barcelona (Room IC) | 🗓️ October 20, 2025 | 🕔 17:00 – 19:45

​As artificial intelligence becomes increasingly capable, ensuring its safety and reliability has become a central challenge for research and industry alike. This event, hosted by AI Safety Barcelona, brings together leading voices from policy and technical research to explore how we can better understand, evaluate, and govern advanced AI systems.

​Agenda

​17:05 – 17:15 | Introduction
Welcome and overview of AI Safety Barcelona’s mission to connect researchers, engineers, and policymakers working toward safer AI development.

​17:15 – 18:00 | Alejandro Tlaie – Understanding AI Risks (Online)
Alejandro Tlaie, AI Policy Advisor at Pour Demain, will discuss current and emerging AI risks, from near-term challenges to long-term governance questions. Drawing on his background in complex systems and philosophy of science, he will outline how the EU and other global actors are approaching risk assessment and regulation, and what opportunities exist for collaboration between research and policy.

​18:10 – 18:55 | Pablo Moreno – Evaluating AI Beyond Benchmarks (Online)
Pablo Moreno, researcher at Xanadu AI and VRAIN (UPValencia), will introduce the General Scales framework, a new paradigm for evaluating AI systems. This approach uncovers what benchmarks truly measure and helps predict model capabilities with greater precision. The talk will highlight how such methods can guide both scientific understanding and responsible deployment in industry.

​19:00 | Coffee & Networking
An opportunity to meet researchers, practitioners, and policymakers shaping the future of AI safety in Europe.

​Join us to explore new ideas, connect with the AI safety community, and engage with the people working to make advanced AI systems truly safe and beneficial.