The institution defines in its internal rules and directives the requirements that ensure the data AI applications use is complete and correct, keeps its integrity, stays available and is accessed only as intended, because AI learns from data without human intervention and data may be wrong, inconsistent, incomplete, unrepresentative, outdated, biased from history or no longer representative after a change in the environment; for purchased solutions it addresses its limited knowledge of and influence over the underlying data and the risk of manipulated data, and it is able to judge the quality of unstructured data such as text and images.
The graph holds this control, the 0 it maps to, and the evidence behind each claim, over MCP and REST.