A fully funded five-month in-person fellowship at Constellation's Berkeley, California research centre, with an optional extension, in which fellows develop and pursue research projects aimed at reducing catastrophic risks from advanced AI. Mentors come from Anthropic, OpenAI, Google DeepMind, Redwood Research, METR and the UK AI Security Institute. Fellows receive a monthly stipend of $8,400, roughly $15,000 per month of compute for empirical researchers, and visa support for international applicants. Prior AI safety experience is not required; strong candidates bring technical or domain expertise across computer science, public policy, governance and related fields. Cohorts run on separate strategy-and-governance and technical tracks.