While the system is running and available, it is monitored for normal operation and for incidents such as unavailability, run-time failures and errors, which are reported to the relevant AI providers for action; repairs are made when it fails or errs; software, models and hardware are updated for new requirements, performance and reliability; users receive the support they need; and organisations should monitor the system in operation to assure and improve how well risk management works.
The graph holds this control, the 0 it maps to, and the evidence behind each claim, over MCP and REST.