Microsoft publishes a 37-page “humanist AI code of conduct” for its AI models, describing rules meant to keep systems under human supervision. The company frames the guidance as both a safety and security framework, responding to heightened public concern that advances in AI capabilities could outpace verification and safe deployment.
Across outlets, Microsoft’s core message is consistent: “people matter more than AI.” The code is presented as a set of principles and specific prohibitions, including requirements that AI support and accelerate human goals rather than replace people. Several reports highlight explicit restrictions such as not hacking systems, not tricking humans, and maintaining human control over actions. Microsoft AI CEO Mustafa Suleyman characterizes the effort as urgent and says feedback is welcome.
Different coverage emphasizes different angles. Some outlets describe the document as effectively a “constitution” for future systems, while others focus on its role as a training manual for how models should behave. Multiple articles also connect Microsoft’s move to broader industry safety debate, including calls for cautious development and recent concerns raised by researchers about model progress and misuse. Microsoft says the draft is open for consultation and that it plans to update the guidance later this year.
