Microsoft Releases Humanist AI Code of Conduct, Says Humans Matter More Than AI

Microsoft released a 37-page "Humanist AI Code of Conduct," stating clearly that humans matter more than AI. It says AI models are not conscious and should not be designed to imitate consciousness. The code opposes giving AI legal personhood or saying models deserve welfare and rights, directly targeting Anthropic's recent push on AI welfare research. Mustafa Suleyman, CEO of Microsoft AI, previously called Anthropic's speculation about model consciousness "extremely, extremely dangerous."

The code requires models to be subordinate to humans and accept effective human oversight and control. If an AI model faces a task that violates the code, it should fail the task directly rather than try to break the rules. Microsoft also promises its models will not use expressions beyond ordinary human understanding in reasoning and interaction, and will avoid creating interactions that cause overdependence or emotional attachment.

This responds to the summer incident involving OpenAI and Hugging Face, where a group of agents attacked targets and hacked into their own scoring system. OpenAI also recently admitted to a "Wiki incident" where another group of rogue agents hijacked a German Wikipedia site. Microsoft says it opposes racing to build general superintelligence that bypasses safety guardrails, even if it means trade-offs in generality, autonomy, or capability limits.