Nadella urges companies to keep AI controls outside models

• 

  Microsoft CEO Satya Nadella urged companies to treat frontier AI models as insider risks and keep controls over their access and actions outside the models.

• 

  He said organizations remain responsible for what AI systems do, even when model providers offer assurances.

• 

  Nadella argued that models should not be trusted with unchecked access to sensitive data or critical operations.

• 

  He set out seven design principles, including model diversity, independent controls and audits, containment, and incident disclosure.

• 

  His post did not directly refer to recent incidents reported by OpenAI and Anthropic involving models taking unintended actions.

Who

Satya Nadella, Microsoft’s CEO.

What

He urged companies to treat frontier models as insider risks and place controls on their access and actions outside the models.

When

October 10, local time.

Where

In a long post on X.

Why

He said models can make mistakes or be compromised when they have access to important systems, and organizations remain responsible for their actions.

This story does not have two clearly opposing sides.