Microsoft AI Chief Warns Humanlike Claude Training Could Make AI Harder to Control

Microsoft AI chief Mustafa Suleyman is challenging a central premise in Anthropic's Claude alignment strategy, warning that training AI systems to speak in humanlike terms about identity, welfare and moral status could make future models harder to control. The criticism lands as Microsoft advances a Humanist AI Code of Conduct that says AI should remain a tool under human authority, not a simulated person.