In model building the team keeps room for human review of outputs during training and tuning, keeps the model secure in an environment with appropriate access levels, designs for diverse conditions and stability, and follows privacy processes. It ensures the model is explainable to the degree fixed at planning, and where explainability is limited weighs benefits against that limit, documents the reasons for proceeding and increases human oversight. It tests continuously for bias in outputs with oversight of purpose, constraints, principles and decisions, considers environmental impact and better choices, assesses effects on wellbeing and society, maintains audit trails and impact assessments, and ensures the body can report on the actions and decisions that led to an outcome and respond to its consequences.
The graph holds this control, the 0 it maps to, and the evidence behind each claim, over MCP and REST.