Short Bio

Charbel-Raphaël Segerie is the Executive Director of CeSIA, the French Center for AI Safety, and an OECD AI Expert. CeSIA is part of the official evaluator consortium for the EU AI Office’s Code of Practice on harmful manipulation risks (2026–2028), having contributed to all three drafting rounds with recommendations included verbatim in the final text.

He initiated and co-led the Global Call for AI Red Lines, endorsed by 12 Nobel laureates and 10 former heads of state, presented at the UN General Assembly by Nobel Peace laureate Maria Ressa, and highlighted by Yoshua Bengio at the UN Security Council. The campaign received 300+ media mentions worldwide (NYT, BBC, Le Monde, TIME, NBC).

In 2022, before ChatGPT, he created at ENS Paris-Saclay and ENS Ulm the first university-accredited course on general-purpose AI safety in the EU — a curriculum covering capabilities scaling, alignment, and misuse and loss-of-control risks that anticipated the concerns raised by frontier models such as Claude Mythos. He founded ML4Good, an EU Commission-funded AI safety programme replicated 20+ times worldwide with a 98% participant recommendation rate and alumni now at the EU AI Office, Mistral, MATS, and many other institutions. He is the scientific director of the AI Safety Atlas, used by 1,000+ students worldwide.

He was previously Head of AI Safety at EffiSciences, and has experience in the industry as a startup CTO (Omniscience) and in French research institutions (Inria Parietal, NeuroSpin/CEA). His research focuses on identifying emerging risks in AI, with peer-reviewed publications at ACM FAccT (co-authored with Yoshua Bengio), ICML, and TMLR on RLHF limitations, LLM safeguards, and AI safety evaluation methods.

He convened workshops at the India AI Impact Summit (Delhi, 2026) and co-organized the AI Safety Symposium (Paris, 2025) with keynotes by Yoshua Bengio and Stuart Russell; He has presented twice at the French Senate on AI risks and governance, appeared on France Inter and many podcasts, and collaborated on a YouTube video on AI risks reaching 5 million views.