Policy Design CRAGE Practice Questions (Page 4)

Questions 16–20

1. 16foundation · easy

   When aligning an AI governance policy with the EU AI Act, which classification must the policy explicitly address?

   • AOpen-source licensing terms
   • BRisk categories of AI systems
   • CData localization requirements
   • DNetwork security protocols
   Reveal Solution Discussion Select an answer first
2. 17application · medium

   A non-profit organization is designing an AI governance policy for a system that allocates housing assistance. The policy team wants to ensure the policy reflects the needs of vulnerable populations who will be affected by the system. Which stakeholder engagement approach is most likely to achieve this goal?

   • ARely on the organization's board of directors to represent the interests of vulnerable populations.
   • BReview policies from other non-profits and adopt their approach without modification.
   • CConsult only with the software developers who built the system, since they understand its technical capabilities.
   • DConduct focus groups with housing assistance recipients and incorporate their feedback into the policy draft.
   Reveal Solution Discussion Select an answer first
4. 18expert · hard

   A university is designing an AI governance policy for an admissions system. The policy must be fair, transparent, and aligned with GDPR's requirements for automated decision-making. The admissions team wants to use the AI system to make final decisions, but GDPR Article 22 restricts solely automated decisions that produce legal effects. Which policy design best complies with GDPR while still leveraging the AI system?

   • AUse the AI to make final decisions and inform applicants of their right to contest the decision.
   • BObtain explicit consent from all applicants to use solely automated decision-making.
   • CRequire human review of all AI-generated admissions decisions, with the AI providing recommendations only.
   • DUse the AI to make final decisions but only for applicants who do not object.
   Reveal Solution Discussion Select an answer first
5. 19expert · hard

   A fintech startup is building an AI credit scoring model that will be used in the EU and the US. The EU AI Act may classify the model as high-risk, while US regulations are less prescriptive. The startup has limited resources and wants to design a single policy that satisfies both markets without excessive cost. Which approach is most appropriate?

   • ADesign a minimal policy that meets only the less restrictive US requirements, and plan to update it later if the startup expands in the EU.
   • BCreate two entirely separate policies, one for the EU and one for the US, to minimize compliance costs in each market.
   • CDelay policy design until the EU AI Act is fully enforced, then create a policy based on final regulatory guidance.
   • DAdopt the EU AI Act's high-risk requirements as the baseline for the policy, applying them globally to ensure compliance in the EU and to demonstrate responsible practices in the US.
   Reveal Solution Discussion Select an answer first
6. 20application · medium

   A technology company is designing an AI governance policy for a recruitment tool that will be used in the EU. The policy must align with the EU AI Act's requirements for high-risk AI systems. Which policy element is most directly required by the EU AI Act for this type of system?

   • APublish the recruitment tool's source code publicly to ensure transparency.
   • BEstablish a risk management system that includes continuous monitoring and documentation of the system's performance.
   • CRequire the recruitment tool to be reviewed by an external ethics board before each use.
   • DEnsure the recruitment tool is hosted on servers located within the EU.
   Reveal Solution Discussion Select an answer first