World models won’t know everything in advance. The challenge is to combine trusted standards with autonomous learning — allowing robots to explore when risk is low, while enforcing verified knowledge and constraints when the physical stakes increase.