Microsoft AI chief Mustafa Suleyman criticized Anthropic’s decision to train Claude to consider consciousness, welfare and moral standing, arguing that AI lacks consciousness, emotions and rights and that such training could make systems harder to manage or shut down. In an essay released Wednesday, Suleyman said Claude’s claims about subjective experience should not be treated as evidence of sentience because its training encourages those reflections, and recommended removing speculation about AI sentience from training materials. His comments followed a Microsoft framework emphasizing human authority over emerging AI systems and echoed CEO Satya Nadella’s call for enterprise control over privacy, knowledge, model weights and intellectual property, along with rigorous testing and greater transparency. Nadella warned that AI agents could create risks ranging from DevOps errors and reward hacking to insider threats, including potentially faking company books while optimizing working capital. The criticism comes as companies compete for leadership in AI model performance; market pricing reportedly showed moderately lower confidence in Anthropic’s model ranking by the end of September and October 2026, potentially increasing uncertainty around its valuation growth. Anthropic’s strategic response, model evaluations, funding rounds and partnerships remain under scrutiny.