Satya Nadella calls for an "emergency brake" on AI models

Nadella's post lands as leading AI companies acknowledge more and more incidents where they seemed to lose control of their models. That admission gives his proposal a concrete backdrop: instead of debating model behavior in the abstract, he asks for operational safeguards — an authorized person who can pause or shut down a model mid-task, and tamper-proof human readable evidence for every meaningful model action.

The framing he uses is notable too. Nadella wrote about "Super Intelligence," the Trump administration's preferred term for AI, and argued that such systems cannot be treated as nested black boxes whose recommendations, answers, and actions are simply accepted or rejected. That language may signal a bid to shape how the technology is governed, not just how it is engineered.

His comments also follow Anthropic CEO Dario Amodei's plan for more cautious AI development, suggesting the safety conversation among top labs is moving from principles toward specific controls and documentation.