Lyptus Research

Lyptus Research is an Australian not-for-profit AI safety research group, focused on evaluating dangerous capabilities in frontier AI models and researching how advanced AI systems shape each other’s values as they evolve.

Their work spans two core areas: evaluating how quickly frontier AI models develop potentially hazardous capabilities (with a current focus on biosecurity and offensive cyber risks), and alignment research into how advanced AI systems shape the values of the models that succeed them. Lyptus shares its work openly on GitHub and Hugging Face.