Reliability is the ability of the system, or an entity within it, to deliver its required functions, under the conditions stated, over a stated period; for an AI system it is the ability to give required predictions, recommendations and decisions consistently and correctly during operation. It is affected at least by robustness, generalisability, consistency and resilience: every input and environment setting that meets the stated criteria is expected to be processed correctly, including inputs that differ from those seen in development. Backup of the system or a component, implementing the same business logic, improves reliability when the primary fails, and reliability supports functional safety where stakeholders require automatic protection against defined failures.
The graph holds this control, the 0 it maps to, and the evidence behind each claim, over MCP and REST.