Lausanne AI Alignment

Lausanne AI Alignment (LAIA), operating today as Safe AI Lausanne (SAIL), is a student-led AI safety organization centered at EPFL in Lausanne, Switzerland. It emerged as a direct spinoff of the first ML4Good (Machine Learning for Good) bootcamp organized by EffiSciences, a French nonprofit focused on AI safety field-building. The group is a commission and working group of EA Lausanne and is legally hosted under EA Geneva, a tax-exempt Swiss association. From its founding in early 2023, the group set out to foster research, education, and community engagement around AI safety at EPFL and beyond. A central figure in its early history was Romain Graux, who co-created the group and served as Vice President from January 2023 to June 2024. He led the implementation of the AI Alignment Curriculum funded by BlueDot Impact's AI Safety Fundamentals program and organized a two-week intensive AI Safety bootcamp in September 2023 for approximately 20 participants, covering transformer architecture, mechanistic interpretability, reinforcement learning, RLHF, and LLM security. The group also hosted a series of Alignment Jam hackathons at EPFL, including an interpretability-focused 48-hour event in April 2023 and a Multi-Agent Safety hackathon in September-October 2023, the latter in partnership with the Cooperative AI Foundation with co-authors from UC Berkeley, Oxford, Cambridge, Harvard, and DeepMind. The organization received a grant from Open Philanthropy (now Coefficient Giving) to support its AI Alignment Bootcamp program. Under the Safe AI Lausanne name, the group continues to run bi-weekly reading groups, introductory fellowships, AI control hackathons, and research projects in collaboration with EPFL labs including the MLO and NLP labs. Members have published research at top-tier venues including NeurIPS and ICML. As of March 2026 the group is actively running events including a Spring 2026 kickoff, an AI Safety Reading Group, an Introduction Fellowship, and an AI Control Hackathon.