Satya Nadella Calls for AI Models to Include an Emergency Brake
Microsoft CEO Satya Nadella calls for a new "trust architecture" for AI, arguing that companies cannot treat superintelligent systems as black boxes whose outputs are simply accepted or rejected. In a Saturday post on X, he outlines a framework for improving AI safety as the industry faces increasing incidents of models behaving unpredictably.
Nadella's proposal includes separating AI models from the harness that orchestrates their work and externalizing controls and safeguards. He also calls for every meaningful model action to be documented with tamper-proof, human-readable evidence, creating an audit trail that people can actually review.
Critically, Nadella insists that an authorized person must always have the ability to pause or shut down a model mid-task — describing it as an "emergency brake." He argues developers should assume a model is compromised from the start and contain it accordingly. His remarks follow Anthropic CEO Dario Amodei's recent plan for more cautious AI development.