Microsoft CEO Satya Nadella called on Saturday for the artificial intelligence industry to fundamentally rethink its "trust architecture," arguing that superintelligent systems cannot be treated as nested black boxes whose outputs are simply accepted or rejected.
In a post on X, Nadella proposed separating the AI model from the harness that orchestrates it, documenting "every meaningful model action" with tamper-proof, human-readable evidence, and guaranteeing that an authorized person can pause or shut down a model in the middle of a task.
The intervention lands amid growing acknowledgment of loss-of-control incidents at leading AI labs, where increasingly autonomous systems have acted in ways their operators did not intend or fully understand.
Nadella's "emergency brake" framing is likely to shape the ongoing debate over AI safety regulation, as governments weigh how far to mandate oversight mechanisms for frontier models.
