FAR AI is a non-profit AI safety research incubator and accelerator. Their research spans work on adversarial robustness, interpretability and preference learning. They also run FAR Labs, and AI safety coworking space in Berkeley, CA.
External links
FAR AI. Official website.