Microsoft has published a code of ten principles for the development and use of AI models. The document emphasizes human oversight, the ability to shut down systems, and the unacceptability of granting them legal personhood.