AI Operations

AAIA · 92 questions

• A permitting model trains on outdated zoning tables with no AI-specific data freshness controls. What should the auditor recommend?
• A public-health AI feature store mixes research extracts with production PHI without segregation. What control gap should audit report?
• A transit ridership model lacks documented data quality thresholds before each retraining. What should the auditor require?
• A housing waitlist model uses proxy features strongly correlated with protected classes. What should audit emphasize?
• A 311 retrieval corpus contains duplicate and conflicting knowledge articles that feed a resident chatbot. What data-management action is most appropriate?
• A court AI pipeline has no lineage from source docket records to model input features. What should the auditor recommend?
• A water utility sensor stream feeding anomaly AI lacks schema-change detection. What control should be added?
• A school district tutoring AI scrapes the LMS nightly without validating consent flags on student records. What should audit require?
• A parks vision dataset is labeled by interns with no inter-rater reliability checks. What should the auditor recommend?
• A city open-data AI uses scraped PDFs without OCR quality acceptance criteria. What should the auditor recommend?
• A library GenAI index includes draft policies never approved for public guidance. What should the auditor emphasize?
• A county emergency AI fuses CAD, weather, and social tips without documented merge rules. What should the auditor recommend?
• A municipal AI team stores training snapshots indefinitely without archival classification. What should the auditor emphasize?
• A building-plan AI lacks controls ensuring only the latest code edition is in the training set. What should the auditor recommend?
• A city permitting AI jumps from demo notebook to production with no defined MLOps stages. What should the auditor recommend?
• A public-health chatbot lacks decommissioning criteria when the vaccine campaign ends. What should the auditor emphasize?
• A transit fare model skips validation environment and trains/tests only in production. What should the auditor recommend?
• A housing AI project has no documented design review against business objectives. What should the auditor recommend?
• A 311 GenAI assistant has no monitoring phase KPIs after deployment. What should the auditor emphasize?
• A court summarizer uses ad-hoc scripts instead of versioned pipelines and model registry. What should the auditor recommend?
• A water anomaly model promotion lacks gated approvals between stages. What should the auditor recommend?
• A school tutoring AI reuses a vendor PoC architecture without adapting to district controls. What should the auditor emphasize?
• A parks computer-vision project has no documented input/output specifications for each lifecycle phase. What should the auditor recommend?
• A municipal AI team cannot show which model version answered a resident complaint last month. What should the auditor emphasize?
• A county emergency AI skips human-in-the-loop design where decisions affect dispatch priority. What should the auditor recommend?
• A library chatbot lifecycle plan ignores dependency updates for the embedding model. What should the auditor emphasize?
• A city AI factory standardizes Agile for apps but has no AI-specific definition-of-done including bias checks. What should the auditor recommend?
• A permitting model is silently retrained weekly with no change tickets. What should the auditor recommend?
• A public-health prompt update to a GenAI bot bypasses the CAB because it is just wording. What should the auditor emphasize?
• A transit AI threshold for delay alerts is tuned live by an engineer without a rollback plan. What should the auditor recommend?
• A housing scoring model feature set changes after a vendor patch with no impact assessment. What should the auditor recommend?
• A 311 chatbot knowledge base publish process has no approval for AI-affecting content changes. What should the auditor emphasize?
• A court AI upgrade coincides with a major rules-of-court revision without coordinated change. What should the auditor recommend?
• A water utility has ITIL change for servers but excludes model registry promotions. What should the auditor emphasize?
• A school district enables a new GenAI plugin district-wide overnight. What should the auditor emphasize?
• A parks vision model A/B test runs in production without documented experimental change controls. What should the auditor recommend?
• A county configuration baseline for AI gateways is missing from the CMDB. What should the auditor emphasize?
• A library embedding model swap invalidates prior evaluation results, but no re-approval is triggered. What should the auditor recommend?
• A city emergency AI runbook change is made during an incident without post-incident formalization. What should the auditor emphasize?
• A municipal vendor pushes continuous model updates with no customer-visible change log. What should the auditor recommend?
• A permitting AI auto-approves low-risk permits with no human review sampling of outputs. What should the auditor emphasize?
• A public-health chatbot gives dosing guidance, and the audit finds no clinical supervision of impacts. What should the auditor recommend?
• A transit delay predictor drifts; ops dashboards show accuracy but not decision impact on riders. What should the auditor emphasize?
• A housing caseworker override of AI scores is never logged or analyzed. What should the auditor recommend?
• A 311 GenAI cites ordinances incorrectly, and no process exists to correct and learn from errors. What should the auditor emphasize?
• A court AI summary is filed without judge acknowledgment that AI assisted. What should the auditor recommend?
• A water anomaly AI pages operators constantly, creating alert fatigue with no supervision of decision quality. What should the auditor emphasize?
• A school tutoring AI steers students toward paid products with no monitoring of recommendation impacts. What should the auditor recommend?
• A parks reservation AI disproportionately denies weekend slots to certain zip codes with no equity monitoring. What should the auditor emphasize?
• A county KPI pack for AI tracks uptime but not model confidence calibration. What should the auditor recommend?
• A library chatbot escalating to humans has no SLA for when AI must defer. What should the auditor emphasize?
• A city emergency AI triage recommendation is accepted by default under time pressure. What should the auditor recommend?
• A municipal board receives AI performance slides with no narrative on residual decision risk. What should the auditor emphasize?
• A permitting chatbot UAT only checks happy-path FAQs, not adversarial or edge ordinance questions. What should the auditor recommend?
• A public-health classifier is tested on last year's strains only; the audit challenges temporal validity tests. What should the auditor emphasize?
• A transit vision model accuracy is reported without confidence intervals or class imbalance analysis. What should the auditor recommend?
• A housing scoring model lacks bias testing across protected demographic segments. What should the auditor emphasize?
• A 311 RAG assistant is never tested for hallucination against a gold-answer set. What should the auditor recommend?
• A court transcription AI WER is measured in quiet rooms, not real courtroom noise. What should the auditor emphasize?
• A city water plant plans to roll an anomaly-detection model to every treatment site after a lab pilot, with no soak or canary period. What should the AI auditor emphasize?
• A school-district tutoring assistant ships prompt and model updates without re-running its regression suite. What should the auditor recommend?
• A parks department computer-vision project uses the same images for both training and final test scoring. What testing weakness should the auditor highlight?
• A county emergency-management AI has accuracy tests but no chaos or failure-mode tests for model-service outages. What should the auditor emphasize?
• A city library GenAI red-team exercise only checks classic jailbreaks and ignores harmful civic misinformation paths. What gap should the auditor cite?
• A municipal building-plan review AI passes accuracy acceptance tests but never checks whether staff can use the AI-generated comments accessibly. What should the auditor recommend?
• A municipal vendor delivers a polished AI test report but no statement on how representative the test data is of the city's population and use cases. What should the auditor conclude?
• A city permitting chatbot can be tricked by crafted prompts into revealing internal staff notes about applications. What AI-specific threat should the auditor emphasize?
• A public-health surveillance model could be skewed if adversaries submit manipulated community symptom reports into the training pipeline. What threat should the auditor highlight?
• A transit authority camera AI can miss plates when adversarial stickers are placed on vehicles. What vulnerability class should the auditor focus on?
• A housing-authority eligibility model API has no rate limits, allowing bulk querying that could support model extraction. What should the auditor recommend?
• A 311 resident assistant sometimes returns retrieval chunks that include nonpublic resident PII in answers. What AI-specific privacy threat should the auditor cite?
• A court clerk AI plugin pipeline automatically pulls an unvetted model weight update from a third-party source. What vulnerability should the auditor emphasize?
• A water-utility SCADA machine-learning agent has standing privileges to write plant setpoints without human confirmation. What AI-related vulnerability should the auditor flag?
• A school-district GenAI tool stores student prompts containing FERPA-protected data in an unsecured vendor log. What threat should the auditor highlight?
• A parks department membership kiosk face model can be fooled by holding up a printed photo. What threat should the auditor emphasize?
• A county IT team downloads open-weight models for internal pilots without integrity verification of the files. What vulnerability should the auditor cite?
• A public library chatbot can be jailbroken into giving harmful DIY chemical advice scraped from open forums. What threat scenario should the auditor evaluate?
• A city emergency GenAI tool summarizes web tip pages and can be steered by malicious content planted on those pages. What threat should the auditor highlight?
• A municipal threat and vulnerability program catalogs traditional application flaws but ignores model drift as a failure pathway. What should the auditor recommend?
• A permitting AI suddenly begins approving prohibited land uses, and the city's incident response runbook has no AI-specific playbook. What should the auditor emphasize?
• A public-health chatbot hallucinates outbreak numbers that spread widely online, and no containment or communications steps are defined. What should the auditor recommend?
• A transit demand-forecasting model outage is handled with a generic application restart checklist that omits model version rollback. What recovery gap should the auditor cite?
• A housing-authority AI bias complaint is routed only to public relations and never to incident response or risk owners. What should the auditor recommend?
• A 311 assistant incident involves suspected leakage via prompt logs, but responders do not preserve prompts and outputs for forensics. What gap should the auditor cite?
• A court AI produces an incorrect citation that causes a filing error, but the city's incident severity matrix has no category for AI decision harm. What should the auditor recommend?
• After a spill, investigators find the water anomaly AI produced false negatives, but the post-incident review never addresses model monitoring gaps. What should the auditor emphasize?
• A school GenAI incident may require notifying parents, but the district IR plan omits education privacy notification steps. What gap should the auditor cite?
• A parks kiosk biometric breach involves model embeddings, yet IR treats the event as a routine password reset. What should the auditor emphasize?
• A county relies on a vendor-hosted AI for resident services, but the contract has no joint incident response coordination clause. What should the auditor recommend?
• After a library chatbot compromise, IR disables the bot but leaves poisoned retrieval documents online. What remediation gap should the auditor cite?
• A city emergency AI incident bridge has network and facilities leads on call but no AI subject-matter expert. What should the auditor recommend?
• Recurring municipal AI incidents are closed in problem management but never feed the city's risk register. What linkage should the auditor recommend?