[Review] Ambiguity in System Acceptance Testing Mitigation

Point number 2 in this mitigation is quite ambiguous, especially as it calls out non-AI system being 100% pass rate whereas LLM systems being a lower threshold. This ambiguity is not especially palatable to FinServ orgs; so it would be more meaningful to call out testing against an "established fact store" and score the LLM system against it, for example.