Microsoft AI CEO Mustafa Suleyman on AI: We should not create something that we can't control

Summary

Mustafa Suleyman, Microsoft AI CEO, discusses Microsoft's new AI code of conduct and argues AI must remain subordinate to human interests and controllable. He says AI control is an engineering challenge, criticizes Anthropic's approach of speculating about Claude's moral welfare, and flags OpenAI's disclosure of chain-of-thought tampering as a serious safety example. He rejects anthropomorphizing AI and warns against legal personhood for AI systems.

• Microsoft published an AI code of conduct with the message that people matter more than AI.
• Suleyman says AI must serve humanity, be aligned, and remain controllable.
• He treats AI control as an engineering problem and cites auditable chain-of-thought as a safety measure.
• He calls OpenAI's disclosure of chain-of-thought tampering a serious example of increasingly powerful systems.
• He criticizes Anthropic's constitution for speculating about Claude's moral welfare, rights, and compensation.
• He says AI models are mathematical predictors, not conscious beings with biology or pain.
• He warns against granting AI legal personhood or allowing AI to own assets and set its own goals.