A Cambridge-based hub bringing together students and professionals to reduce existential risks from advanced AI systems through education, research mentorship, and community-building.
A Cambridge-based hub bringing together students and professionals to reduce existential risks from advanced AI systems through education, research mentorship, and community-building.
People
Updated 05/18/26Funding Details
Updated 05/18/26- Annual Budget
- -
- Current Runway
- -
- Funding Goal
- -
- Funding Raised to Date
- -
Org Details
Updated 05/18/26Cambridge AI Safety Hub (CAISH) was founded in September 2022 in Cambridge, UK, with the aim of supporting people in Cambridge and around the world who want to contribute to addressing societal-scale risks from advanced AI. The organization started with a single full-time organizer, Hannah Erlebach, who ran the hub until summer 2023. After a leadership transition supported through the Open Philanthropy University Organiser Fellowship, the hub continued under new leadership and has since expanded its programs and community significantly. CAISH currently operates three main programs. The Alignment Fellowship is a free, five-week in-person course held each year in Cambridge, offering parallel tracks for participants with technical backgrounds (covering model training, interpretability, and AI control) and governance backgrounds (covering policy, regulation, and geopolitical strategy). The Alignment Desk is a structured writing and research accountability program for community members with existing AI safety exposure who want to produce and publish serious work. MARS (Mentorship for Alignment Researchers) is a part-time research mentorship program, founded in 2024 by former CAISH director Chloe Li, that pairs teams of two to four participants with experienced mentors from organizations such as Redwood Research, MIRI, UK AISI, and Stanford for 10-12 weeks of remote research following an in-person kick-off week in Cambridge. MARS provides participants with travel funding, accommodation, office space, catering, and over $2,000 in compute credits at no cost. CAISH is legally constituted as a project of Meridian Impact CIC (UK Companies House number 13653958, formerly Cambridge Effective Altruism CIC), a Community Interest Company headquartered at 53-54 Sidney Street in central Cambridge. This arrangement gives CAISH access to Meridian's co-working space, meeting rooms, and broader network of researchers working on AI safety, biosecurity, and related cause areas. The hub is currently co-directed by Justin Dollman and Gaurav Yadav, with support from a student committee and a Director of Events and Operations. CAISH has received operating cost funding from Open Philanthropy (grant announced November 2025). Alumni and community members have gone on to organizations including the UK AI Safety Institute, DeepMind, GovAI, Apollo Research, MATS, and Anthropic.
Theory of Change
Updated 05/18/26CAISH believes that the AI safety field is talent-constrained and that a key bottleneck is the pipeline of researchers and practitioners who deeply understand both the technical and governance dimensions of the problem. By embedding itself in Cambridge's research ecosystem and running high-quality, free educational and mentorship programs, CAISH aims to identify and develop talented individuals who would not otherwise enter AI safety work, accelerating their transition into productive roles at frontier labs, policy organizations, and academic research groups. The MARS program specifically targets the gap between introductory education and independent research by pairing promising students with mentors who can guide them to publishable, field-relevant work, thereby growing both the supply of safety researchers and the body of useful research.
Grants Received
Updated 05/18/26Projects
Updated 05/18/26A six-week intensive "AI Safety Fast-Track" programme run by Cambridge AI Safety Hub, combining courses and workshops to introduce students to key ideas in AI safety.
A structured writing programme and AI safety "writing accelerator" for people in Cambridge with prior AI safety exposure who want to produce and publish serious work.
A condensed five-week in-person course in Cambridge with technical and governance tracks that introduces motivated students and working professionals to fundamental concepts in AI safety.
Discussion
No comments yet. Be the first to share your thoughts.