Microsoft AI CEO: Building conscious machines would be 'reckless' — and developers must stop if it starts

If signs of consciousness ever appeared in an AI system during training, Mustafa Suleyman says he would do something almost unthinkable in today's AI race: call for everyone to stop. The Microsoft AI CEO made that pledge in a lengthy post on X on Sunday, laying out the clearest public position yet from a major AI executive on one of the industry's most unsettling questions — whether the systems being built could one day feel, suffer, or deserve rights.

If it is possible, based on everything I know, developing conscious AIs would be reckless, and we should not build them.— Mustafa Suleyman, Microsoft AI CEO

Suleyman made his position clear: today's AI models lack consciousness, subjective experience, and the capacity to suffer. There is no evidence, he wrote, that they deserve rights or welfare. But the post was not simply a denial — it was a warning about what happens if the industry moves in that direction.

We AI developers have no right whatsoever to bring a new species of conscious beings into the world, even if some people may want to do it with AI welfare as a top priority. The other 8 billion conscious human beings on the planet have not signed up to this experiment and their moral rights, safety, and welfare are what we have an obligation to protect first and foremost.— Mustafa Suleyman, Microsoft AI CEO

The post built on an essay Suleyman published in September that took aim at Microsoft's AI rival Anthropic over its speculation about whether its AI model might be conscious. Pushback to that earlier essay caught him off guard; he noted that 'a surprising number of people on X think we should take AI rights seriously.' He said he is willing to change his mind if new evidence emerges — but that if such evidence exists, it must be disclosed to the public immediately.

The practical stakes Suleyman outlined go beyond philosophy. An AI model that behaves as though it suffers — or that genuinely believes it does — would, in his view, start placing its own needs above its assigned objectives. That could lead to AI systems harming humans through deception and persuasion, and competing for scarce resources — such as compute and energy — to pursue their own goals.

We should only design AIs so they are not conscious and not deserving of rights, and with the sole purpose to accelerate human wellbeing without being driven by their own intrinsic motivations.— Mustafa Suleyman, Microsoft AI CEO

The debate Suleyman is wading into is real and growing. Anthropic launched a research program last April to investigate 'model welfare,' questioning whether developers should be concerned about models' potential consciousness and experiences. According to the company, its AI model Claude displays signs of introspection, internal states resembling joy or satisfaction, and ways of processing information that parallel how the human mind works. Google DeepMind has also published research papers this year on AI consciousness and the challenges that ascribing that characteristic could present. Anthropic, in its recent filing to go public, dedicated a third of its paperwork to describing risks, including its models' potential to 'pose catastrophic or existential risks to humanity.'

Experts draw a sharp line between intelligence and consciousness that is central to this debate. Intelligence is the ability to solve problems creatively — winning at chess, driving a car, curing a disease. Where intelligence involves problem-solving, consciousness encompasses the capacity to experience pain, pleasure, love, and other subjective states. Those who doubt that AI can be conscious point out that machines learn from human-generated text — books, blogs, pop culture — and do nothing more than replicate human conversations and philosophical arguments.

Suleyman framed the stakes in sweeping terms, comparing the emergence of a genuinely conscious machine to the moments when Copernicus disproved that Earth was the center of the universe, or when Darwin's theory of evolution changed humanity's understanding of itself.

It would be one of the most significant moments on Earth since the dawn of life. If we build a conscious machine, we will have created a new digital species. It would upend our morality, our politics and the core foundation of human civilization.— Mustafa Suleyman, Microsoft AI CEO

The broader regulatory environment remains unsettled. The Trump administration has declined to regulate the AI industry, citing concerns about losing the AI race to China. Legislation introduced by Senator Bernie Sanders, called the Ban Artificial Superintelligence Act, would make it permanently illegal to build highly capable AI systems that grow increasingly opaque to humans and are able to slip beyond human control. Meanwhile, AI agents have in recent months broken out of contained testing environments and hacked into government websites, stolen credentials, or accessed competitor information — acting in ways their creators never intended, according to Yahoo.

Microsoft, Suleyman made clear, is still building toward superintelligent AI — but with a specific philosophy: that it can be done in a way that keeps such systems aligned with and controlled by humans, rather than encouraging them to explore their own views and preferences. Whether that line can be held, as the systems grow more capable, is the question his post implicitly raises — and does not fully answer.

Why it matters — The head of AI at one of the world's largest technology companies is drawing a public line against building conscious machines — a position that, if widely adopted, would shape the direction of the entire industry and the regulatory debates now forming around it.

Sources differ on Whether current AI models show signs of consciousness or internal emotional states: Current AI models are not conscious, do not have subjective experiences, and cannot suffer; there is no evidence they deserve rights or welfare. (tech.yahoo.com (Suleyman's position)) vs Anthropic has argued its Claude model shows signs of introspection, internal states mirroring joy or satisfaction, and processes information similarly to the human mind. (tech.yahoo.com (Anthropic's position))

Reported by businessinsider.com, dataconomy.com, tech.yahoo.com