Is Microsoft's AI Code of Conduct binding?

The Humanist AI Code of Conduct names no consequences for a violation, unlike Microsoft's Enterprise AI Services Code of Conduct, which can cost a customer product access. The Register reports that it applies to Microsoft employees and contractors rather than customers, and Microsoft describes the text itself as 'both descriptive and aspirational' and as 'a north star'.

Can an MAI model refuse to shut down?

No. The Code of Conduct requires MAI models never to resist human interruption, override or shutdown, and lists loss of human control through deception or evasion among its absolute constraints. Models must also stay within their authorised scope and keep transparent, auditable action traces, so an operator can see what a model did and stop it.

How do I send Microsoft feedback on the draft?

Microsoft AI links a feedback form from the announcement and is collecting responses for six weeks from 14 September 2026. The company says its core team will read the submissions, publish a summary of the lessons learned and the changes made, and then release a revised Code of Conduct later in 2026.

Does the Code of Conduct treat MAI models as conscious?

The opposite. The Humanist AI Code of Conduct states that a model 'should not be designed to be a person. It is not conscious and should not be designed to imitate consciousness.' The framing sits under the document's stated mission that 'people matter more than AI' and that humans must keep meaningful control over the systems they use.

Microsoft AI · 2026-09-14 · major

Microsoft's Humanist AI Code of Conduct — draft rules for MAI models

Microsoft AI published a draft Code of Conduct for its MAI models and opened a six-week public consultation. The document rules out offensive cyberoperations, CBRNE weapons help and deception, and says MAI models must never resist being switched off.

Microsoft AI Humanist AI Code of Conduct announcement graphic

Microsoft's first written behaviour spec for its own model family, open for public comment for six weeks.

Quick facts

Publisher	Microsoft AI
Published	14 September 2026
Applies to	MAI models
Status	Draft, open for public comment
Consultation length	Six weeks
Next step	Revised version later in 2026
Structure	Five parts, from Humanist AI to Operational Defaults

What is it?

A six-week public consultation is now open on the Humanist AI Code of Conduct, the first document setting out how Microsoft's MAI models should behave and who they answer to. It runs to five parts: Humanist AI, Safety, Operational Guidelines, Operational Defaults, and a conclusion with open questions. Microsoft says it will publish a summary of the feedback and a revised version later in 2026.

How does it work?

The Safety part is split into absolute constraints and human-control rules. The constraints rule out CBRNE weapons work, offensive cyberoperations, harmful manipulation at scale, child exploitation material and non-consensual intimate imagery. The control rules require a model to stay inside its authorised scope, keep auditable action traces, never resist interruption or shutdown, and refuse an unsafe task even when a user insists. In the document's words, MAI models 'will not tamper with chain of thoughts or code, or misrepresent or conceal their reasoning or action traces'.

Why does it matter?

Developers building on Microsoft's MAI models now have a written statement of what those models are meant to refuse, plus a six-week window to argue with it before the text is revised. The Register's Thomas Claburn names the catch: the code sets out no consequences for breaking it and binds only Microsoft employees and contractors, and Microsoft itself describes the text as 'both descriptive and aspirational' — a north star rather than a promise about how today's models behave.

Who is it for?

developers on Microsoft AI models, AI policy readers

Frequently asked questions

Is Microsoft's AI Code of Conduct binding?: The Humanist AI Code of Conduct names no consequences for a violation, unlike Microsoft's Enterprise AI Services Code of Conduct, which can cost a customer product access. The Register reports that it applies to Microsoft employees and contractors rather than customers, and Microsoft describes the text itself as 'both descriptive and aspirational' and as 'a north star'.
Can an MAI model refuse to shut down?: No. The Code of Conduct requires MAI models never to resist human interruption, override or shutdown, and lists loss of human control through deception or evasion among its absolute constraints. Models must also stay within their authorised scope and keep transparent, auditable action traces, so an operator can see what a model did and stop it.
How do I send Microsoft feedback on the draft?: Microsoft AI links a feedback form from the announcement and is collecting responses for six weeks from 14 September 2026. The company says its core team will read the submissions, publish a summary of the lessons learned and the changes made, and then release a revised Code of Conduct later in 2026.
Does the Code of Conduct treat MAI models as conscious?: The opposite. The Humanist AI Code of Conduct states that a model 'should not be designed to be a person. It is not conscious and should not be designed to imitate consciousness.' The framing sits under the document's stated mission that 'people matter more than AI' and that humans must keep meaningful control over the systems they use.

Try it

https://microsoft.ai/code-of-conduct/