Microsoft AI’s Humanist AI Code of Conduct – Keeping Humans in Control

Microsoft AI has published the first draft of its Humanist AI Code of Conduct, and I thought this was worth sharing as AI continues to become more capable and agentic.

The idea behind it is quite simple — AI should work for people and remain under meaningful human control.

The Code outlines how Microsoft wants its own MAI models to behave and the boundaries it expects them to operate within.

Some of the areas that caught my attention include:

• Humans remain in control – AI should allow human intervention, correction and shutdown.

• AI remains artificial – it shouldn't claim to be conscious, have feelings or present itself as a human.

• Clear safety boundaries – certain safeguards should remain in place regardless of how the model is configured.

• Support human agency – AI should help people understand, decide, create and get things done rather than unnecessarily taking control away from them.

I think this becomes particularly interesting as we move from AI assistants towards agents that can reason, access information, use tools and take actions.

For enterprises, the conversation is therefore no longer only about what AI can generate. We increasingly need to think about what AI should be allowed to do, where human approval is required, and how identity, permissions, governance and accountability fit around it.

One important point is that this is currently a draft Code specifically for models developed by Microsoft AI (MAI). It shouldn't be interpreted as a policy covering every Microsoft AI product or every model used within Microsoft's ecosystem.

Microsoft has opened the draft for public consultation for six weeks and plans to refine it before using the Code to help guide future MAI model development.

If you are following Responsible AI, AI governance or the evolution of enterprise agents, this is worth a quick read.

Read more:

Microsoft AI – Humanist AI Code of Conduct
https://microsoft.ai/code-of-conduct/

Microsoft AI – Humanist AI in Practice: Public Consultation on the Code of Conduct
https://microsoft.ai/news/mai-code-of-conduct/