Microsoft CEO Satya Nadella has called for a rethink of how trust is built into AI systems, becoming the latest technology leader to set out detailed views on improving AI safety. In a Saturday morning post on X, Nadella said it was time to step back and assess the trust architecture underpinning AI.
Using the Trump administration’s preferred term, Super Intelligence, Nadella argued that such systems cannot be treated as a series of nested black boxes whose recommendations, answers and actions are simply accepted or rejected. Instead, he proposed separating AI models from the harness that orchestrates their work and moving controls and safeguards outside the model itself.
Nadella also said every meaningful action a model takes should be recorded with tamper-proof, human-readable evidence, and that systems should always allow an authorized person to pause or shut down a model in the middle of a task. He said developers should assume a model has been compromised and contain it from the outset, likening the approach to an emergency brake.
His comments follow a series of incidents in which leading AI companies acknowledged apparently losing control of their models, and come after Anthropic CEO Dario Amodei published a plan for more cautious AI development.
