A water utility compliance matrix assigns who signs off on AI safety evidence before release. What is being established?
Select an answer to reveal the explanation.
Short Explanation
You need a named signer for the safety packet—roles and skills beat just hoping someone’s available. Annual safety training refreshers don’t assign sign-off authority, and versioned evidence repositories don’t name who reviews AI safety packets either.
Full Explanation
A water utility compliance matrix that assigns who signs off on AI safety evidence before release establishes defined compliance KSAs and sign-off roles for civic AI safety evidence. Named authority and required skills prevent vague “someone should check” culture and make release accountability auditable for systems that can affect water-quality communications, customer guidance, or operational decision support.
Hiring external law firms to review red-team findings addresses legal liability after the fact, but it doesn't establish who inside the utility is accountable for signing off before release, which is what the matrix is for. Coordinating annual safety training refreshers for utility operations personnel is about workforce training, not about naming who signs off on AI safety evidence; it doesn't establish the compliance KSAs or sign-off roles the matrix requires. Maintaining versioned repositories of model evaluation evidence for audit readiness supports documentation, but it doesn't assign named sign-off authority or define the specific KSAs needed to review AI safety evidence before release. The matrix's job is naming who, with what expertise, signs off on what evidence.
Exam caveat: a matrix without trained, available signatories is paper compliance. Roles must match actual expertise in safety, security, privacy, and operations, and must escalate when evidence packages are incomplete or contradictory.
Operational check: publish the matrix with role, required KSA, artifact checklist, and signature or date fields; verify a release cannot proceed in the change system without completed AI safety sign-offs from the assigned authorities. Train backup signatories so release does not stall when the primary reviewer is unavailable during a maintenance window.