Theo Farrell

First-class MSci Natural Sciences (Computer Science & Philosophy) from Durham University. Founder and co-organiser of Durham AI Safety Initiative. Previously I worked on relational composition and SAEs (i.e. mechinterp) but now I’m more interested in model organisms, emergent misalignment and inoculation prompting. I’m also increasingly concerned about risks from non-frontier open weight models. I’m currently in London for LASR Labs, investigating evaluation of model organisms (mentored by Satvik Golechha). Outside of AI Safety I play bass guitar 🎸 and dabble with piano and viola 🎹🎻!