An AI system is engineered to produce outputs (content, forecasts, recommendations, decisions) for objectives that humans define; it has no understanding of its own, so human design choices, engineering work and oversight are needed. How much oversight depends on the use case, but at a minimum it is present during training and validation, and it serves to keep development and use aligned with intent and to keep stakeholder impacts under consideration across the life cycle.
The graph holds this control, the 0 it maps to, and the evidence behind each claim, over MCP and REST.