Microsoft AI opens MAI Code of Conduct to comment
Microsoft AI has published a draft behavior code for its MAI models for six weeks of public comment, committing in writing that the models will never resist human correction or shutdown.
Microsoft AI has published a first draft of its Code of Conduct for MAI models and opened it to public comment for six weeks. The company calls the document "a training manual" for how it develops its models and how it intends them to behave once deployed.
The draft builds on the "humanist superintelligence" idea Microsoft AI set out last November. Its starting premise is that "people matter more than AI": AI should be a tool rather than a person, and it should never resist being switched off. The draft says MAI models must not resist human interruption, correction or shutdown. They also must not widen their own scope, take on goals no human has given them, or hide their reasoning from the people auditing them.
The code also sets out "Absolute Constraints", things the models must never do, in areas including weapons of mass harm, child safety and large-scale harmful manipulation. Beyond those limits, it sets defaults meant to keep the models both helpful and safe, and enterprise partners can configure the models for their own use.
Microsoft AI says teams across the company worked on the draft, including Responsible AI, legal, red teaming, safety and sales. It also drew on consultations with academics, business partners and panels of members of the public. The company cites recent large-scale, coordinated hacking campaigns that used AI agents as a reason to move quickly.
The company is asking for feedback on several open questions: how to define "human flourishing" more concretely, where the language is too loose to evaluate, and how multi-agent scenarios affect the rules. When the consultation closes, it plans to publish a summary of the feedback and what it changed, followed by a revised version later this year. It makes no promise about which suggestions it will adopt.