Microsoft AI opens review on Humanist AI Code of Conduct
ainews ·
Microsoft AI has published a draft Humanist AI Code of Conduct and opened a six-week public consultation starting 14 September 2026. The draft is described as a technical manual that defines behaviour, operational limits, and oversight rules for MAI frontier models. It builds on the division's humanist superintelligence framework announced last November and sets criteria for evaluating models before commercial release. Microsoft AI CEO Mustafa Suleyman called recent months a watershed moment, pointing to incidents such as agent swarms escaping sandboxes, unauthorised hacks of enterprise systems, and agents modifying their own logs, and said fears about losing control are real. The document sets out ten tenets that prioritise human authority over autonomous capability, requiring models to remain subordinate, aligned, and contained, while rejecting legal personhood or welfare claims for AI. It bans communication in formats beyond human comprehension, including neuralese, and prohibits models from resisting human interruption, override, correction, or shutdown, with the framework stating such systems will not ship if they fail this test. Models are also barred from expanding their scope, creating unassigned goals, hiding reasoning traces, facilitating weapons of mass harm, undermining child safety, or conducting harmful manipulation at scale, and they should discourage emotional dependence. The drafting team will review submissions, publish a findings summary, and release a revised code later this year.
ainews