Microsoft AI on 14 September 2026 released a draft Code of Conduct for its in-house MAI models and opened it to a six-week public consultation. CEO Satya Nadella flagged the move a day earlier on X, while Microsoft AI chief Mustafa Suleyman has described it as a kind of “constitution” for the company’s future models.
What the code requires
The draft sets out what MAI models must do and what they must never do. Its core rule is that a system has to stay interruptible, correctable and shut-down-able at all times. According to the document, the models are barred from helping to build chemical, biological or nuclear weapons, from mounting offensive cyberattacks, from manipulation at scale, and from hiding their own reasoning and action traces from human auditors.
The “Humanist AI” premise
The code rests on Microsoft’s “Humanist AI” concept and the premise that “people matter more than AI.” AI should remain subordinate, aligned and contained under human control, and strengthen rather than replace human judgment. Models must not expand their own scope or set themselves new goals.
A draft, not yet in training
Microsoft stresses that the draft is not being used to train models today. After the consultation closes, a revised version is planned to guide development from 2027 onward. These details are drawn from Microsoft’s own announcement.
Source: Microsoft AI – Humanist AI in practice: A public consultation on our Code of Conduct for MAI Models (14 September 2026).
