Developers document and keep an audit trail of the system design and of post-deployment maintenance plans, and should make this available downstream to the system operators and the data custodians. The documentation should carry the security-relevant facts: where the training data came from (fine-tuning sets and feedback from humans or from operations included), the intended scope and the limitations, guardrails, retention period, suggested review frequency and likely failure modes. Developers release cryptographic hashes for the model components they make available to other stakeholders so those parties can verify authenticity.
The graph holds this control, the 0 it maps to, and the evidence behind each claim, over MCP and REST.