The provider must detect users' input with keyword, classifier and similar methods, and must set and publish to users a rule under which a user who repeatedly enters illegal or unhealthy information in succession, or reaches a set number of such inputs in a day, has the service suspended or faces other handling measures; it must also appoint monitoring personnel, in numbers matched to the scale of the service, who improve the quality and security of generated content in light of what they observe, track national policy and collect and analyse third-party complaints.
The graph holds this control, the 0 it maps to, and the evidence behind each claim, over MCP and REST.