Microsoft CEO Satya Nadella wants AI agents to do far more on our computers. There’s just one small catch: He doesn’t think we should trust them.

“I think of all agents as insider risk and treat them as such — or all models as insider risk — and so treat them as such,” Nadella told Joanna Stern in an exclusive interview following Microsoft’s unveiling of its Nvidia-powered Surface Laptop Ultra.

It’s a striking admission from the head of a company that’s not only building AI models but also makes an operating system used by more than a billion people — the very place where those potentially risky AI agents will operate.

But that’s also central to Nadella’s vision.

Subscribe to NBC News to read the full article.