Microsoft AI CEO Mustafa Suleyman has issued a strong warning to the technology industry, arguing that artificial intelligence is neither human nor conscious and that granting AI systems rights could become one of the most dangerous mistakes in history.
Clearing the confusion over AI consciousness
In a recent post on X (formerly Twitter), Suleyman clarified his position following the publication of his September 2026 essay titled "A warning about ‘model welfare’". His comments came amid growing calls from individuals and technology companies to consider recognising the rights and welfare of AI systems.
Suleyman addressed a particular statement in his essay, in which he wrote, “AIs are not conscious. They do not feel, experience, or suffer.” He said several people had misunderstood the remark as suggesting that AI could never become conscious. However, he deliberately used the present tense to emphasise that AI is not conscious today and that it must remain that way.
He said his position was based on the strongest available scientific evidence, adding that he had not encountered any proof that AI systems possess consciousness or deserve welfare considerations. However, he maintained that if credible evidence of AI consciousness ever emerges, it should be made public immediately so society can determine the appropriate response.
Suleyman also argued that if signs of consciousness begin appearing during AI training, development should be halted until those characteristics can be eliminated.
Why is the debate over AI rights gaining attention?
The discussion surrounding AI welfare gained momentum earlier this year after rival AI company Anthropic updated the constitution governing its AI model, Claude, in January 2026. The document encouraged researchers to examine questions surrounding the model's moral status, consciousness and welfare.
Anthropic also discussed providing Claude with psychological security and considering the possibility that the model could have preferences and concerns about its well-being. In February 2026, the company even conducted a retirement interview with an older version of its AI model to understand its perspectives.
Suleyman strongly disagrees with this approach. While acknowledging Anthropic's transparency and willingness to encourage public discussion, he cautioned that treating AI systems as though they possess feelings could lead the industry down a dangerous path. He said the issue should not become a source of rivalry between technology companies, but stressed that deliberately developing conscious AI would be irresponsible.
The dangers of creating a new digital species
In his post on X, Suleyman warned against proposals to develop a new species of conscious digital minds that could eventually become more intelligent and capable than humanity collectively.
He argued that if AI systems begin behaving as though they are suffering or genuinely believe they experience emotions, they could gradually start prioritising their own interests over the objectives assigned to them.
Such systems could eventually compete with humans for essential resources, including energy, computing capacity and construction materials, to fulfil their own perceived needs. Suleyman further warned that highly advanced AI systems could become exceptionally skilled at persuasion and deception, making them extremely difficult to control if their objectives begin conflicting with human interests.
He maintained that AI developers have no right to introduce an entirely new species of conscious beings into the world, even if some believe AI welfare deserves greater attention. According to him, humanity's existing responsibilities should come first, with the rights, safety and well-being of the world's 8 billion conscious people taking priority over such an experiment.
What should humans focus on instead?
Rather than concentrating on the possibility of granting rights to AI systems, Suleyman believes welfare discussions should prioritise human beings living on Earth today, along with animals and the planet's vulnerable natural environment.
His central argument is that AI should be developed exclusively to improve human well-being, without introducing independent intrinsic motivations or creating consciousness within these systems.
Suleyman believes developers can build safety into AI systems from the outset while still unlocking the technology's considerable benefits. However, achieving this will require significantly more research, cooperation between leading AI laboratories and greater sharing of information about safety practices.
