
Most organizations can prove they have an AI policy. Far fewer can prove what actually happened the last time an AI system shaped a real decision.

Trustworthy AI cannot be reduced to model behavior. It depends on trustworthy systems of oversight around AI.