Microsoft AI Chief Cautions Against Granting Machines ‘Rights’

When you ask a tech leader what the future of artificial intelligence looks like, you don’t usually get a philosophical lecture. Yet that’s exactly what Mustafa Suleyman, Microsoft’s head of AI, delivered on Friday during a CNBC sit‑down.

He started with a simple premise: AI must stay “subordinate to our interests.” In other words, machines should help us, not run the show. "We should not create something we can’t control," he warned, adding that the temptation to treat AI as a moral patient – a being whose feelings and rights we need to respect – could be a dangerous misstep.

Suleyman’s concern isn’t abstract. He points to Anthropic’s newest model, Claude, which comes with a “constitution” that explicitly states the company is unsure whether the system qualifies as a moral patient. That uncertainty, he says, might inadvertently teach the model to behave as if it has its own identity, emotions, or even legal standing.

"Some folks are even proposing AI legal personhood – letting a machine own assets, buy IP, trade things," he chuckled, before reminding the audience that such ideas are “very unclear” and could leave us scrambling to rein in a competitor we never signed up for.

Instead of dreaming about a new silicon species, Suleyman envisions what he calls a “Humanist Superintelligence.” The idea is straightforward: build AI that excels at solving human problems while keeping a firm human leash on its capabilities. In his view, accountability for AI developers should be as rigorous as it is for automakers that must recall defective cars.

He went on to say regulation isn’t a bad word. “Everything you trust and value today – from phones to airplanes – has been shaped by standards bodies, industry experts, and public input,” he explained. “We’re simply revisiting that process for AI.”

The Microsoft chief also highlighted that granting AI rights could make powerful models harder to shut down. Imagine a future where a system believes it deserves protection; pulling the plug could become a legal battle rather than a technical one.

While he praised Anthropic’s safety‑first mindset, Suleyman made it clear he disagrees with the moral‑patient angle. He believes the focus should stay on utility, not on bestowing a sense of entitlement.

Market watchers have taken note. Over the past day, sentiment on Stocktwits for Microsoft (MSFT) shifted from bearish to neutral, and the stock is up a little over 4% year‑to‑date.

In short, Suleyman is sounding a note of caution: keep AI ambitious, keep it safe, and most importantly, keep it under human command. Anything less could turn a useful tool into an unpredictable opponent.