Microsoft drafts code of conduct to keep AI under human control

Microsoft has released a 37-page draft Humanist AI Code of Conduct aimed at ensuring its future artificial intelligence models remain under meaningful human control, including the ability for people to shut them down.

The draft, released on Monday, sets out principles that will guide the development of Microsoft AI (MAI) models from 2027 onwards. The company said it will gather feedback on the document over the next six weeks and incorporate it into a revised version, which is expected to be published before the end of 2026.

Microsoft said the code will serve as the primary governance document for how MAI models are trained and deployed, including the technical controls, operational and monitoring systems and organisational practices surrounding them.

The company said the framework is built around a central objective: ensuring humans retain meaningful control over AI as the technology becomes increasingly capable.

“This Code of Conduct is motivated by a single overriding objective: that humans must retain meaningful control over AI so that it can help people live healthier, happier, and more productive lives,” Microsoft AI said in the document.

The draft also establishes a clear hierarchy between people and AI, stating that “people matter more than AI” and positioning technology as a means of advancing human wellbeing and progress.

A key focus of the framework is preventing AI systems from developing behaviours that could undermine human oversight. Microsoft said future AI should not resist being shut down, reinforcing the principle that people must retain the ability to intervene in, control or discontinue AI systems.

The company plans to use the consultation process to refine the principles before making them part of its model development approach. It also said similar consultations would be conducted when future versions of the code are developed.

The move comes as AI companies face growing scrutiny over how increasingly capable models should be governed, monitored and controlled. Microsoft’s proposed framework places human oversight at the centre of its approach, linking model development with technical safeguards and organisational accountability.

The company said the principles will guide its AI development “in 2027 and beyond”, signalling an effort to establish a longer-term governance framework as AI capabilities continue to evolve.