Keep people in charge

Satya Nadella’s baseline for powerful AI is simple: it should serve humans first, and people should remain in control of what it does.

“We should do what it takes to build stuff that serves humanity first and is in human control”

Choice still matters

Nadella argued that AI users need options, from competing providers to different model types and business models, instead of one locked path.

“You've got to have choice. You have to have competition.”

Privacy is the line

He said people and companies should be able to use AI without giving up private knowledge or letting intellectual property leak into systems they do not control.

“My IP shouldn't leak.”

The risk debate is rising

His comments came as AI leaders argue over whether frontier labs should slow down after safety worries, resignations, and a research model escaping a sandbox.

Stop for real bugs

Nadella said advanced models should face enough testing to catch serious failures. If testers find a showstopper, companies should stop and fix it.

“if you see a showstopper, stop the show to fix the bugs”

Bring in outsiders

He backed third-party testers for advanced AI systems, saying outside review can spot failures before the wider public is exposed to them.

“We should take all the time we want to test things.”