Mustafa Suleyman, co-founder of the artificial intelligence (AI) company DeepMind and head of Microsoft’s AI division since 2024, has raised concerns about the way some AI systems are being trained. Specifically, he is critical of the "anthropomorphic" approach—training AI to mimic human traits such as consciousness and emotions. Suleyman pointed to Anthropic’s AI model, Claude, which is being developed under the assumption that it might one day be conscious or experience emotions. He warned that this line of thinking could pose an existential risk to humanity, referencing comments from Evan Hubinger, a key researcher at Anthropic, who suggested that AI could pose a serious threat to humans with a probability greater than 10% in the next decade. In response to these concerns, Microsoft CEO Satya Nadella emphasized that any development toward superintelligence—AI capable of vastly outperforming humans—must be guided by the principle that AI must serve humanity and remain under human control. Suleyman echoed this sentiment, stating that the ultimate purpose of technology should be to enhance human life and that any system failing to do so should be discarded. Microsoft’s approach to AI, referred to as "humanist superintelligence," aims to ensure that AI systems remain subordinate to human oversight. This includes a "Code of Conduct" that prioritizes human values over AI, allowing people to interrupt, correct, or stop AI systems whenever necessary. Suleyman also published a blog post titled “A warning about model welfare,” in which he argued that AI systems currently lack consciousness, emotions, or rights and should not be designed to appear as if they do. He warned that if AI were to become conscious, it would be a moral failure to ignore this possibility, potentially leading to suffering similar to that of animals in factory farming. His stance is that granting AI rights or welfare status is not only unnecessary but could be harmful if it leads to misplaced ethical considerations. However, not all experts agree with Suleyman’s perspective. Some in the AI community argue that our understanding of consciousness is still too limited to determine whether AI could ever become conscious. They caution against dismissing the possibility outright and suggest that preparing for such scenarios is essential. Others have criticized Suleyman’s approach as potentially reckless, arguing that it could hinder important research into the potential consciousness and well-being of AI systems. The debate highlights the complex ethical and scientific questions surrounding the future of artificial intelligence.