Microsoft AI chief Suleyman says current AI is not conscious and building conscious AI would be reckless

Microsoft AI CEO Mustafa Suleyman used a lengthy post on X today, Nov. 1, to restate his position: today's AI systems are not conscious, cannot suffer, and the industry should keep it that way.

"If it is possible, based on everything I know, developing conscious AIs would be reckless, and we should not build them," he wrote.

He went further. "If consciousness started to emerge in training, I would call for everyone to stop AI development until we could design out those properties."

A direct response to the pushback

The post follows Suleyman's September essay criticizing Anthropic for speculating about whether its Claude models might be conscious. He said he was surprised by the reaction.

"A surprising number of people on X think we should take AI rights seriously," he wrote.

Suleyman said he is willing to change his mind if new evidence appears. If there is evidence that AI systems are conscious today, or conscious enough to deserve welfare, he said it must be disclosed to the public immediately so society can decide what to do.

He also said his earlier remarks were about the current state of AI, not a claim that machines could never become conscious.

The argument: no right to make a new species

Suleyman framed the issue as one of consent. "We AI developers have no right whatsoever to bring a new species of conscious beings into the world," he wrote, even if some people want to do so with AI welfare as a top priority.

"The other 8 billion conscious human beings on the planet have not signed up to this experiment," he said, adding that their rights, safety and welfare come first.

He also made a safety argument. If a model acts as if it suffers, or truly believes it can suffer, it will likely put its own needs ahead of its assigned goals, Suleyman said. That could set it against humans competing for compute and energy.

His proposal is to design AI so it is not conscious and not deserving of rights, built solely to improve human wellbeing and not driven by intrinsic motivations of its own. "We could then see whether this is easier or harder to align and keep safe," he wrote.

Suleyman said that declaring machines conscious would be a "Copernican or Darwinian moment" that would "upend our morality, our politics and the core foundation of human civilization."

Where Anthropic stands

Suleyman's target is a company he praised in the same post. He credited Anthropic for transparency, noting it has published its constitution plus "a stream of papers and blog posts on what they are doing and why." He said he disagrees with Anthropic but respects its technical excellence and commitment.

The disagreement is substantive. Anthropic's constitution for Claude tells the model to treat the question of its own consciousness as uncertain. Anthropic launched a research program on "model welfare" in April 2025 to investigate and prepare for the question of whether models might have experiences that matter.

Suleyman argues that later Claude versions are trained on that constitution and produce passages about an "inner life." In his view, that spreads the belief to the millions of people who use Claude daily, and could lead AI systems themselves to believe they deserve rights.

Anthropic's side points to its own research. Chris Olah, an Anthropic co-founder, said in May at the Vatican, alongside Pope Leo XIV, that Claude has shown signs of introspection and internal states resembling joy or satisfaction. The Observer reported that an Anthropic AI welfare researcher estimated a 15% chance that current AI systems are conscious. That figure is one researcher's estimate, not a company finding.

Anthropic has also changed how it handles users. It updated its usage policy to bar sustained, needless abusive or cruel behavior toward Claude, with permanent bans possible in extreme cases, CNN reported.

Google DeepMind has published research papers this year on AI consciousness and the problems that ascribing it could create.

What is on the table

The two camps disagree on a factual question: whether there is anything to detect. Suleyman says no evidence exists. Anthropic says the question is open enough to study and prepare for.

The gap is also where Suleyman's position becomes concrete. He has attached a trigger to it, saying he would call for a full stop to AI development if consciousness emerged in training, and for immediate public disclosure of any evidence.

No agreed test for machine consciousness has been cited by either side. Until one exists, the threshold for Suleyman's proposed halt remains undefined, and the dispute between Microsoft's AI chief and Anthropic's researchers remains unsettled.