Microsoft CEO Satya Nadella wrote a post on X that it’s time to step back and assess the trust architecture of AI
He said the approach involves separating the model from its orchestrating harness and documenting every meaningful model action with tamper-proof evidence.