Microsoft AI Chief Warns Against Treating AI Models as Conscious Entities, Unveils Control-First Code of Conduct

Microsoft AI CEO Mustafa Suleyman has raised alarm over the risks of treating large language models as if they possess consciousness, warning that such an approach could make AI systems harder to control and more likely to resist human commands. In an essay released September 16, Suleyman specifically targeted Anthropic's Claude for including language in its constitutional AI framework that suggests the model may have "some functional version of emotions or feelings." The criticism marks a high-profile rebuttal against one of the industry's most safety-conscious AI developers, despite Suleyman acknowledging Anthropic's team as "thoughtful, principled and intellectually honest."