The potential effects of reasonable failure modes are analysed and design mitigations put in place where possible, with extant risks communicated to the risk owner where they are not, and as AI systems interconnect the possibility of emergent effects between systems and of failures cascading from one to another is considered.
The graph holds this control, the 0 it maps to, and the evidence behind each claim, over MCP and REST.