A transit AI report omits disclosure that vendor model internals were inaccessible during testing. What reporting element is missing?
Select an answer to reveal the explanation.
Short Explanation
If you could not see inside the vendor model, say so in the report. Hidden limitations make transit AI assurance look stronger than it was. A summary of tests on accessible parts isn't disclosure because it leaves out the inaccessible internals.
Full Explanation
Disclosure of scope and evidence limitations is missing when a transit AI report omits that vendor model internals were inaccessible during testing. Readers will otherwise overestimate the depth of assurance obtained over black-box components that drive rider-facing predictions. Transparent limitation language is a core reporting element for municipal AI audits that rely on third-party systems the city cannot fully inspect. Stating what could not be tested protects governance from false confidence and frames residual risk honestly.
Providing only a summary of testing procedures performed on accessible system components fails because it describes what was examined without disclosing the critical limitation that vendor model internals were inaccessible, so it does not supply the missing scope-and-evidence-limitation disclosure. A statement of auditor independence and objectivity for the engagement fails because it addresses the auditor's stance rather than the omitted disclosure that vendor model internals were inaccessible, so it does not supply the missing scope-and-evidence-limitation reporting element. A description of remediation timelines agreed with the transit vendor team fails because it concerns future corrective scheduling rather than the omitted disclosure that vendor model internals were inaccessible during testing, so it does not supply the missing scope-and-evidence-limitation element.
Exam caveat: limitation disclosure does not excuse omitting feasible alternative procedures such as output testing, monitoring review, or contractual assurance reports; both alternatives performed and residual limits should appear. Operational check: record inaccessible artifacts in the transit AI scope memo, describe alternative tests executed, quantify what residual uncertainty remains, and place clear limitation wording in the report's scope or methodology section before issuance.