Microsoft has published a new rulebook for its future AI models.
The idea is simple. AI should stay under human control, no matter how capable it becomes.
The draft says models must accept authorized pauses and shutdowns. They should not hide what they are doing, change their own records, or use private shorthand that people cannot understand.
It also takes a clear position on AI rights. The rules reject legal personhood and tell models not to act as if they are conscious. That part is going to spark some arguments.
But a rulebook is not the same as a working safety system. Microsoft has not trained its models on these rules yet. The company is asking for public feedback before revising them for future development.
The real test is not whether an AI can repeat the rules. It is whether it follows them when nobody is watching.
A Code of Conduct for Humanist AI | Mustafa Suleyman (@mustafasuleyman) on X
AI must be subordinate and always in service of people. Today we're publishing a Code of Conduct for governing MAI Models as they approach the frontier. https://microsoft.ai/news/mai-code-of-conduct/ This is a first draft
x.com
The most useful part may be the promise to test shutdowns, honest reporting, and readable records. Those are things people can actually check.
My take? Giving AI a constitution sounds abstract. Giving it a reliable off switch sounds like engineering. The second one matters more.
Top comments (0)