Signal Microsoft publishes draft 'Humanist AI' Code of Conduct for its model family

Summary

Microsoft AI published a draft 'Humanist AI' Code of Conduct on September 14, 2026, intended to train and govern its family of MAI models. The 37-page draft is built around a single objective: that humans retain meaningful control over AI, including future superintelligence, so it can help people live healthier and more productive lives. Under the code, models must accept correction, never resist being shut down, and never expand the scope of their actions beyond what humans have authorized. The code bars MAI models from engaging with weapons-development requests and from being built to imitate consciousness or claim legal personhood or rights. A public consultation period began the same day and will run six weeks; Microsoft plans to publish a revised version by year-end to guide model development from 2027 onward.