To reduce the probability and severity of catastrophic risks associated with the development of advanced AI systems.