MAF paragraphs 3.36 to 3.44: keep decisions and the data and processes behind them traceable through an audit trail of training and AI-augmented decisions, black-box recorders of input streams, and protected retention suited to the industry, scoped by severity of harm, maturity of the model and regulatory needs; support reproducibility by an independent team through context testing, verification methods, replication files and checking with vendors of off-the-shelf systems; prepare for internal or external audit by keeping records of where data came from, how it was procured and pre-processed, its lineage and how it is stored and secured, ideally in a central process log, without having to publish business models or intellectual property.
The graph holds this control, the 0 it maps to, and the evidence behind each claim, over MCP and REST.