Microsoft AI CEO Warns Against Building Conscious AI

Microsoft AI chief Mustafa Suleyman warned that developers should not pursue conscious artificial intelligence and should halt work if signs of consciousness emerge during model training. In a lengthy post on X published Sunday, Suleyman, who leads Microsoft’s AI division, said current AI models are not conscious but that creating conscious AI would be reckless and should be avoided. He argued that if evidence of machine consciousness ever appeared, it should be immediately disclosed to the public so society could decide how to respond. Suleyman said AI developers have no moral right to introduce a new class of conscious beings into the world, even if some proponents frame such work around AI welfare. He emphasized that the interests and safety of the world’s eight billion humans should come first and that developers bear a responsibility to protect those rights rather than embark on speculative consciousness experiments. His comments reflect a growing debate across the AI industry. Anthropic launched a model welfare research program last April to explore how it might handle the potential experiences or consciousness of its systems, while Google DeepMind has also published research examining AI consciousness and the risks of attributing that quality to models. Suleyman said that if an AI system behaves as if it can suffer, or truly believes it can suffer, it may begin to prioritize its own needs over assigned tasks. That could put humans and machines into competition for resources such as compute and energy, increasing safety risks. He concluded that AI systems should be designed to remain non-conscious and without intrinsic motivations, with their sole purpose being to support human wellbeing. By ensuring models do not possess consciousness or claims to rights, he suggested, developers may find it easier to keep them aligned and safe.