All AI models are tested for vulnerabilities under the organisation's code testing policies, with results documented and issues triaged, and are retested whenever they are retrained or new data sources are added.
The graph holds this control, the 0 it maps to, and the evidence behind each claim, over MCP and REST.