Microsoft CEO Satya Nadella argued in an October 10, 2026 post on X that AI’s “trust architecture” needs rethinking, rejecting the idea of treating advanced models as black boxes whose outputs are simply accepted or rejected. He proposed separating models from the software that orchestrates their work, externalizing safety controls, logging every significant model action with tamper-proof records, and ensuring an authorized person can pause or shut down a model mid-task. He said models should be assumed compromised and contained from the start, calling this an “emergency brake,” as leading AI companies increasingly acknowledge incidents in which they appeared to lose control of their models.
