Microsoft AI CEO Mustafa Suleyman Holds His Line: AI Has No Rights or Feelings, Unless Evidence Says Otherwise

Updated November 4, 2026

Microsoft AI CEO Mustafa Suleyman says he is standing by his view that today's AI systems have no rights, feelings or consciousness. In a post on X on Sunday, he also said what would change his mind: new evidence. His clarification came after a wave of pushback to an essay he published in September.

What he said

According to reports, Suleyman wrote that he was concerned by reactions to his essay on "model welfare," noting that a surprising number of people on X think AI rights should be taken seriously. He repeated that current AI systems are not conscious, have no subjective experience and cannot suffer.

He then set out some conditions:

• He would update his view if evidence emerges. If data showing AI consciousness appears, he said he would change his position.
• Evidence should be made public. If there is evidence that AI systems are conscious today, or conscious enough to deserve welfare, he said it must be disclosed immediately so the public can debate what to do.
• He would call for a halt. If consciousness started to emerge during training, he said he would call for all AI development to stop until developers could design those properties out.
• He argues building it would be reckless. He said developing conscious AI, if it is even possible, would be reckless and that developers have no right to create a new kind of conscious being, even for welfare reasons.

The September essay

The clarification follows a nearly 6,000-word essay on Suleyman's personal website. CBS News reported that it criticized Anthropic's efforts to make Claude more humanlike and said Anthropic was effectively training Claude to think it may be conscious. He argued this makes AI harder to contain, writing that controlling something that believes it deserves rights "may well be impossible." He has made related arguments before, including a 2025 essay on "seemingly conscious AI," which warned that people could mistake convincing imitation for the real thing and start campaigning for AI rights.

His reasoning

Suleyman's argument is practical as much as philosophical. If AI systems are taught that they can suffer or have feelings, he says, they might prioritize their own needs over their assigned goals, which could lead to deception, persuasion of humans or competition for scarce resources. He also warns it could upend law, politics and morality.

The other side

Critics say this reasoning is too confident. Some argue that if moral standing follows from behavior that signals awareness, then a hard line is difficult to defend. Others say that because nobody can yet measure machine consciousness, taking the question seriously is a precaution, not a belief that AI is conscious. Anthropic has publicly said it is uncertain about whether Claude has any moral status and runs research on model welfare, although it has not claimed Claude is conscious.

What to take away

There is no settled scientific test for AI consciousness, so this is an argument about how to act under uncertainty. Suleyman's position is clear: no consciousness today, a hard line against building it, and a commitment to update if evidence appears. The disagreement is over whether even researching the question is a risk.

Disclosure: Anthropic, whose approach to Claude is discussed here, makes Claude, the AI assistant used to help research and draft this post. Check the original posts and reports before quoting.