Oxford AI Safety Initiative

The Oxford AI Safety Initiative (OAISI) is a community of students and researchers at the University of Oxford committed to reducing societal risks from advanced artificial intelligence. OAISI was founded in Summer 2024 when the former OxAI Safety and Governance team decided to spin out into an independent organisation dedicated specifically to supporting AI safety research and education at Oxford. It operates as a project of Jericho Alignment Infrastructure, a company limited by guarantee registered in England and Wales (company number 15988754, incorporated October 1, 2024).

OAISI's work spans both technical and governance dimensions of AI safety, reflecting a belief that AI safety is a sociotechnical issue requiring expertise across multiple domains. The organisation runs a range of recurring programmes: ARBOx is a two-week intensive ML safety bootcamp that has run multiple times and covers building GPT-2-small, interpretability techniques, RLHF, and replicating key research papers. The AI Strategy Series is a multi-part workshop on strategic issues in AI safety and governance. The Governance Roundtable is a weekly discussion group on frontier AI developments and policy implications. Technical Roundtables are weekly researcher gatherings for sharing developments in AGI safety. Office Hours provide informal access to AI safety guidance for students.

OAISI also runs supervised research labs that pair students with experienced supervisors for technical AI safety projects, a Strategy Fellowship for building AI safety metastrategy skills, and has hosted events such as a Narrating AI Futures symposium. The group launched from within the broader Oxford AI Society (OxAI) before becoming independent, a model the founders found effective for reaching non-EA students with STEM backgrounds. Graduates of OAISI programmes have gone on to work in AI safety organisations.

The committee as of late 2025 includes James Lester (President, MPhil Economics, Balliol), Juliana Eberschlag (Programme Manager), Matthew Farrugia-Roberts (DPhil Computer Science, Magdalen), Louis Thomson (CS and Philosophy), and Rohan Selva-Radov (Advisor).