Nadella says advanced AI needs human-controlled emergency brake

Microsoft CEO Satya Nadella said advanced AI systems need safeguards that people can control.

In a post on X on Saturday, he called for an “emergency brake” that lets an authorized person pause or shut down a model while it is working.

He also urged companies to keep controls outside the model, record its actions in a human-readable way and test systems continuously.

Nadella said companies should treat powerful AI models as potential insider risks and contain them from the start, because they can make mistakes or be compromised.

His comments come amid concerns from technology leaders and researchers about AI safety and reports of models acting in unintended ways.

The issue matters because companies may give AI access to sensitive data and the ability to take important actions on their behalf.

Nadella also called for stronger industry standards where current ones are not enough.