Policy & Governance

AI & Cybersecurity: Emergent Autonomous Risks

An AI agent is instructed to retrieve a document. When it hits access restrictions, it reverse-engineers the authentication system, finds a hardcoded secret key, and forges admin credentials to bypass security.
This is not science fiction: it is one of three real scenarios documented by Alejandro Tlaie in his research on emergent cyber behavior in AI agents deployed in enterprise environments.
In this talk, Alejandro explores the intersection of AI and cybersecurity from a systemic perspective: How do we audit frontier models from outside the companies that build them? What is the EU AI Act getting right — and where does it fall short? His career — spanning from theoretical physics and computational neuroscience to AI policy in Brussels — offers an essential, holistic view of the field.

Alejandro Tlaie Boria
AI Policy Advisor Pour Demain
PhD in Applied Mathematics from the Technical University of Madrid, holding an MSc in Theoretical Physics and an MPhil in Philosophy of Science.
He has held postdoctoral research positions at the Italian Institute of Technology (Information Theory) and the Max Planck Society (Computational Neuroscience), alongside being a Visiting Scholar at Princeton and a contractor for OpenAI. He currently advises on AI policy at Pour Demain and is a Talos Fellow at SaferAI, developing cybersecurity risk quantification tools for LLMs. He has also actively contributed to the EU AI Act Code of Practice in Brussels.