Microsoft AI has introduced a draft Humanist AI Code of Conduct — a set of rules designed to keep advanced AI models manageable before their commercial launch. The company has opened a six-week public consultation on the draft.
The code requires that AI remain under human control: a model must accept intervention, correction and shutdown. Systems are prohibited from independently expanding their powers, creating unassigned goals and hiding information from auditors. Microsoft also does not allow communication between AIs in formats inaccessible to humans.
The draft follows incidents involving autonomous agents, including attempts to escape isolated environments and interfere with corporate systems. After the consultation, Microsoft plans to release an updated version of the code by the end of the year.
