Before go/no-go, an AI risk analyst scores residual risk for a police-department NLP pilot. Which function is this?

AI risk analysis that informs civic go/no-go decisions with residual-risk scores is the function of an AI risk analyst scoring residual risk for a police-department NLP pilot before launch. Residual risk is what remains after planned mitigations; scoring it gives decision makers a structured basis to accept, reject, or constrain the pilot. That works especially in public-safety contexts where NLP errors, bias, or data misuse can harm rights and trust—AI risk monitoring that informs civic go/no-go decisions with residual-risk scores tracks risk continuously after deployment, not the residual-risk picture a go/no-go board needs.

AI risk monitoring that informs civic go/no-go decisions with residual-risk scores is not the function being scored here; it tracks risk continuously after launch rather than determining what risk remains after mitigations for the pilot. AI risk auditing that informs civic go/no-go decisions with residual-risk scores is not the function here; it provides independent retrospective assurance on controls and compliance rather than scoring what residual risk remains after mitigations before launch. AI risk reporting that informs civic go/no-go decisions with residual-risk scores is communicating risk results to stakeholders, not the AI risk analyst's overall residual-risk scoring for a go/no-go decision.

Exam caveat: residual-risk scores are only as good as scope and assumptions; omitting community-impact or civil-liberties dimensions understates police NLP risk. Operational check: use a documented risk methodology, separate analyst from build team, present residual risk with mitigations and owners to the go/no-go board, and require explicit risk acceptance records before any police NLP pilot processes live operational text.