Implement AI Solutions Responsibly AI-102 Practice Questions (Page 5)

Questions 21–25

1. 21expert · hard

   A global social media platform uses Azure AI Content Safety to moderate text and images. The platform has a list of banned terms that must be blocked in all regions, but some terms are only banned in specific countries. The moderation team also wants to prevent users from using homoglyphs (look-alike characters) to bypass the filter. What is the most effective approach?

   • AUse a single global blocklist that includes all terms from all regions.
   • BCreate multiple custom blocklists, one for global terms and one for each region, and apply the appropriate list based on the user's location, with variation matching enabled on all lists.
   • CUse the standard content filters only, since they are trained to detect homoglyphs.
   • DDisable text moderation and rely on image moderation to catch all banned content.
   Reveal Solution Discussion Select an answer first
2. 22expert · hard

   A government agency is deploying an AI system to screen public comments for offensive content. The agency must ensure that the system is fair and does not disproportionately flag comments from certain demographic groups. The agency also needs to be able to explain why a comment was flagged. What should you do?

   • AUse a simple keyword-based filter to avoid bias.
   • BUse a third-party AI service that does not provide monitoring or explanations.
   • CDisable all content filters and rely on human moderators to review all comments.
   • DUse responsible AI insights to monitor fairness metrics across demographic groups and generate explanations for flagged comments, while using content filters to block offensive content.
   Reveal Solution Discussion Select an answer first
4. 23expert · hard

   A non-profit organization is deploying an Azure OpenAI-based chatbot to provide mental health support. The organization must ensure that the chatbot does not provide medical advice, and they need to monitor for any signs of harmful behavior (e.g., encouraging self-harm). They also have a limited budget and cannot afford a dedicated team for continuous monitoring. What should they do?

   • AUse groundedness detection with approved mental health resources, enable content filters for self-harm, and set up automated alerts for high-severity events.
   • BSet up a Power Automate flow that emails the organization's staff every time the chatbot is used.
   • CUse a custom blocklist of all medical terms and set the action to 'Reject'.
   • DEnable prompt shields and rely on the chatbot's built-in safety features.
   Reveal Solution Discussion Select an answer first
5. 24expert · hard

   A technology company is deploying an Azure OpenAI-based coding assistant that suggests code snippets. The security team is concerned that the assistant could be prompted to generate malicious code (e.g., SQL injection). They also want to ensure that the assistant does not suggest code that uses deprecated libraries. The company has a responsible AI governance board that meets monthly. What should they implement?

   • AEnable prompt shields to block malicious code-generation prompts, and use groundedness detection with a database of approved libraries.
   • BSet up Azure Policy to deny access to the OpenAI endpoint for any user who has ever triggered a security alert.
   • CAdd common malicious code patterns to a custom blocklist and set the action to 'Reject'.
   • DUse content filters with a low severity threshold for all categories.
   Reveal Solution Discussion Select an answer first
6. 25expert · hard

   A large healthcare provider is deploying an Azure OpenAI-based clinical decision support tool. The provider must comply with regulations that require continuous monitoring of AI safety incidents and a formal process for updating AI governance policies. They also need to ensure that the AI does not provide medical advice outside its approved scope. Which combination of actions should they take?

   • ACreate a custom blocklist of all medical terms and use Azure Monitor to log all API calls.
   • BEnable content filters for all categories and set the severity threshold to 'High'.
   • CUse Azure Policy to enforce tagging on all AI resources and enable prompt shields.
   • DSet up Responsible AI insights with continuous monitoring, establish a quarterly governance review board, and use groundedness detection with approved medical sources.
   Reveal Solution Discussion Select an answer first