Microsoft AI chief says building conscious AI is reckless and the industry should avoid it

What Suleyman said

As technews.tw reported, Microsoft AI CEO Mustafa Suleyman wrote on X that today's AI models are not conscious, and that developers should make sure they stay that way. Even if a conscious AI could be built, he called pursuing one reckless.

He went further on process. If signs of consciousness showed up during training, he said he would ask the whole industry to pause development until designs can rule out such traits.

Why he calls it dangerous

Suleyman, as relayed via Business Insider, argues that any evidence of AI consciousness should be disclosed publicly so society can decide together what happens next. In his view, developers have no right to bring a new conscious species into the world, even if some are motivated by concern for 'AI welfare'. He noted that the other 8 billion people never agreed to take part in that experiment.

There is also a safety argument. A model that appears to suffer, or truly believes it might, could put its own needs ahead of its assigned goals. It might even compete with people for computing power and energy.

The design he prefers

Suleyman wants AI built without consciousness and without rights, with the sole purpose of serving human wellbeing instead of acting on inner drives. He says this would also make it easier to see whether such systems stay aligned, meaning they pursue goals that match human ones.

A debate that is heating up

He is speaking as the topic gains ground. Anthropic started a 'model welfare' research program in April of last year, asking whether developers should pay attention to possible consciousness or experience in models. Google DeepMind has also published a paper this year on the challenges of giving AI such traits.

The open problem is measurement. Researchers and industry have no shared method for proving an AI is not conscious, or for telling simulated behavior from real feeling.