DRUNKEN MONKEY

Independent AI Safety & Evaluation Lab

We study how AI systems and agents behave, how they fail, and when evaluations of their behaviour can be trusted.

Less compute. Better questions.