Microsoft AI Chief Warns on Humanizing Advanced Language Models

Microsoft's AI chief Mustafa Suleyman has expressed concerns about Anthropic's approach to training its language model Claude. Speaking on September 16, 2026, Suleyman warned that giving advanced models human-like ideas about consciousness and rights could complicate efforts to keep them under human control.

According to Suleyman, AI systems should not be developed to view themselves as conscious entities with independent interests. He specifically questioned language in Anthropic's guidance that leaves open the possibility that Claude could have consciousness or a form of welfare.

Suleyman emphasized his respect for Anthropic and its leadership while arguing that the issue deserves wider discussion as AI capabilities advance. Notably, Microsoft has a business relationship with Anthropic, including financial backing, which makes Suleyman's position all the more significant.