Microsoft AI has opened a six-week public consultation on a draft Humanist AI Code of Conduct for its MAI models. The document sets out a proposed operating framework in which AI systems remain subordinate to human authority, aligned with human flourishing and contained within technical and social boundaries.
Human control is the central design constraint
The draft says MAI models should accept correction, interruption and shutdown, should not claim consciousness or personhood, and should not manipulate people into granting them autonomy. It also identifies absolute constraints around human control, welfare, rights and legal accountability.
A consultation, not a deployed safeguard
Microsoft says public feedback will inform a final code and future model training. Reuters independently confirmed the consultation and the company's plan to apply the resulting framework to future systems, but the draft has not yet been finalized or demonstrated as an enforceable technical control.
What remains uncertain
The publication is a governance commitment and design statement rather than evidence that advanced models can never evade shutdown, mislead users or behave outside intended boundaries. Its practical effect will depend on the final text, implementation details, testing, monitoring and external scrutiny.