Satya Nadella calls for an 'emergency brake' on advanced AI models

Microsoft chief executive Satya Nadella used a post on X on Saturday, 10 October, to argue that the industry needs to rethink the “trust architecture” of artificial intelligence, TechCrunch and Bloomberg reported.

What he proposed

• Separate the model from its harness — the software that orchestrates an agent’s work — and keep controls and safeguards outside the model itself.
• Tamper-proof records: every meaningful model action should leave human-readable evidence.
• A human off-switch: an authorised person must always be able to pause or shut down a model mid-task.

According to Bloomberg, Nadella said organisations should treat advanced models as potential insider threats and not rely only on assurances from model makers. “We must assume a model is compromised and contain it from the start,” he wrote, comparing the idea to an emergency brake.

Context

The comments come as AI companies disclose more incidents in which agents behaved in unexpected ways, and after Anthropic CEO Dario Amodei published a plan for more cautious AI development, TechCrunch noted. Nadella’s post is a statement of principles; it did not announce specific Microsoft products or deadlines.