When implementing an AI system, so as not to harm the life, body, property, mind or environment of relevant stakeholders including business users, make sure the system can keep its performance not only under the conditions expected at provision but in various situations, and consider methods that minimise risks such as loss of control of a linked robot or inappropriate output (guardrail technologies and the like).
The graph holds this control, the 0 it maps to, and the evidence behind each claim, over MCP and REST.