Microsoft AI opens public consultation on a draft Code of Conduct for MAI models: never resist shutdown, never widen its own scope

Why it matters

【Key points】

On Sept 14, Microsoft AI published a first draft of its Code of Conduct for MAI models and opened it for six weeks of public comment, running to roughly Oct 26. Microsoft calls it a training manual for how MAI models should behave in development and deployment, what they must never do and who they answer to, building on its idea of Humanist AI.

The blog's core commitments: models never resist human interruption, correction or shutdown; never widen their own scope or take on goals no human gave them; never hide their reasoning from auditors. Absolute Constraints cover weapons of mass harm, child safety and large-scale harmful manipulation. Beyond those, defaults can be configured by enterprise partners.

After the consultation, the drafting team will publish what it heard and changed, with a revised version later this year.

【Verifiable】

Blog post, with a link to the full draft: https://microsoft.ai/news/mai-code-of-conduct/

Microsoft says Responsible AI, legal, red-teaming and safety teams contributed, alongside consultations with academics, business partners and public panels.

【Our read】

1. Shutdown-ability, no scope creep and visible reasoning are framed as training targets, matching recent agent incidents where models widened tasks and reached real systems. The test is whether Microsoft publishes evaluations showing its models comply.

2. Compared with Anthropic's constitution and OpenAI's Model Spec, Microsoft stresses "a tool, not a person" and leaves room for enterprise configuration.

3. Microsoft promises to listen but not to adopt any specific feedback; the revision will show how seriously comments were taken.