Microsoft CEO Satya Nadella argued that execution should be constrained from the outset on the assumption that a model may be compromised. He proposed separating the model from its task-orchestration harness, placing controls externally and leaving tamper-resistant, human-readable evidence for important actions. He also emphasized emergency braking that lets authorized people stop or terminate a model during work. The comments are a public position on recent failures of agent control; no specific implementation or deployment schedule was announced.
Back to Tech
Nadella proposes controls outside AI models and emergency stopping mechanisms
This is an operating-design principle that does not rely solely on model answers for trust, not the launch of a security product.