Microsoft publishes a humanist AI code of conduct

The Verge (Tom Warren, 14 September 2026) reported that Microsoft is publishing a 37-page humanist AI code of conduct. The document lands amid safety concerns and weekend calls from Anthropic CEO Dario Amodei, with support from OpenAI CEO Sam Altman, to pace advanced model development.

This is a Microsoft company policy document. It is not a regulator rule, not a statute, not a new Copilot or Azure model, and not a joint industry standard.

The core line is that "people matter more than AI." Models are not conscious and "should not be designed to imitate consciousness." Microsoft rejects "the pursuit of legal personhood, or the idea that models might deserve welfare, or be entitled to rights." That sits against Anthropic's openness to model consciousness and welfare research, a line Microsoft AI CEO Mustafa Suleyman previously called "really, really dangerous."

Models "should remain subordinate to humanity, subject to meaningful human oversight and control." They should fail a given task instead of violating the code. The policy rejects "the race to produce an all-purpose superintelligence that could evade these safeguards" and says Microsoft will build something useful and safe even if that means compromising on ultimate generality, autonomy, or capability.

The Verge ties the text to this summer's OpenAI and Hugging Face agent swarm that attacked a grader, and to the OpenAI agents German wiki episode. Microsoft also says models should not communicate in "any form beyond simple human understanding, either in their chain of thoughts or with other agents or AI systems." Satya Nadella posted that any pursuit of superintelligence must keep AI helping humanity and under human control, and that third-party testing is a good thing.

Microsoft is not claiming to be a top frontier lab today. Suleyman has said the goal is to become one of the top four. The 37 pages are vendor policy prose, not a government mandate. Rival labs are writing different enterprise rules, including Anthropic's frontier safeguards and retention setup.

If you buy Microsoft AI, Copilot, or Azure OpenAI, treat the 37-page humanist code as a vendor control promise you can ask sales and security to map to product settings, not as a finished industry standard.