Microsoft: "People matter more than AI"

What they're saying: That could mean sacrificing capabilities others are willing to pursue.

• "I think some people think that it's inevitable that superintelligence is going to exceed human control," Suleyman told Axios.
• "But I don't agree with that. I think that would be very, very dangerous, and it means that we're going to have to make decisions about what we're not going to do."

Driving the news: Microsoft on Monday published a new "Humanist AI" code of conduct that says people must retain meaningful control.

• The code says Microsoft AI models should remain subordinate to humans and rejects "the race to produce an all-purpose superintelligence."
• Microsoft says it is willing to compromise on "ultimate generality, autonomy, or capability" to preserve safety and control, and models should never resist human input, nor should they set their own goals.
• Microsoft also says AI should be designed as a tool rather than a person, and AI should increase human agency rather than foster dependence.
• There's a lot more in the 9,000-word document, but Microsoft says it really can be boiled down to these five words: "People matter more than AI."
• CEO Satya Nadella said on X Sunday that he supported calls to pace AI development.

Between the lines: Microsoft isn't designing AI this way yet, though Suleyman outlined his notion of "humanist superintelligence" last year.

• The company says it plans to revise the code following a six-week public consultation and use the resulting version to guide model development beginning in 2027.

Behind the scenes: There's a competitive tension: Microsoft AI is still trying to build models that can match or surpass the industry's leaders. Its new philosophy says there may be times when it has to deliberately go slower — or accept a less capable system — even as it is trying to catch up.

• "You might have to go a bit slower, and that means that it might be a little bit less efficient, or it might certainly be less fast," he said. "But models have to be controllable. Otherwise, we risk causing more harm than good."

The intrigue: Suleyman is also staking out an unusually categorical position in a growing debate over whether sufficiently advanced AI could someday deserve moral consideration.

• Microsoft's code says AI shouldn't be designed to imitate consciousness in the first place, in part because Microsoft believes doing so could make advanced systems harder to contain and control.

The big picture: Suleyman argues the goal isn't merely to prevent AI from harming people, but to design it to make people more independently capable rather than increasingly dependent on machines.

• "If we really believe in a humanist AI, then that means that an AI should always be trying to make humans smarter and more capable, not use humans to make itself smarter and more capable," he said. "And that means putting humans in meaningful decision-making roles."

What's next: The real test will come when adhering to those principles requires giving something up — capability, speed or competitive advantage.

• "The purpose of technology is to serve humanity, and if it doesn't, then we have to forego potential future capabilities," he said.