Design Responsible AI, Security, Governance, Risk Management, and Compliance AB-100 Practice Questions (Page 6)

Questions 26–30

1. 26expert · hard

   A financial firm uses an AI agent that accesses a proprietary model and a grounding database. The firm wants to allow data scientists to update the grounding database but not access the model's weights. The model is stored in Azure Machine Learning. What access control design should you implement?

   • AGrant data scientists read/write access to the grounding database and reader access to the model registry.
   • BGrant data scientists owner access to the entire resource group.
   • CProvide data scientists with the model's endpoint key so they can test it.
   • DGrant data scientists contributor access to the Azure Machine Learning workspace.
   Reveal Solution Discussion Select an answer first
2. 27foundation · easy

   What does data residency compliance primarily ensure?

   • AThat data is accessible only to authorized users
   • BThat data is stored and processed within the geographic boundaries required by regulations
   • CThat data is encrypted at rest and in transit
   • DThat data is backed up daily
   Reveal Solution Discussion Select an answer first
4. 28application · medium

   A global e-commerce company is deploying an AI agent that handles customer service inquiries. The company operates in multiple jurisdictions, including the EU, where data protection laws require that personal data of EU customers be processed only within the EU. The agent uses a central Azure OpenAI service in the US. What should the architect do to comply with data residency requirements?

   • AObtain consent from EU customers to process their data in the US.
   • BDeploy a separate Azure OpenAI instance in an EU region and route EU customer traffic to that instance, ensuring no EU personal data is sent to the US.
   • CUse the US-based Azure OpenAI service but encrypt all data in transit and at rest.
   • DUse a third-party AI service located in the US that promises to delete data after processing.
   Reveal Solution Discussion Select an answer first
5. 29foundation · easy

   Which component is essential for an effective agent governance framework?

   • AFaster inference hardware
   • BA larger training dataset for the agent
   • CClear definition of roles and responsibilities for agent oversight
   • DA user-friendly chat interface
   Reveal Solution Discussion Select an answer first
6. 30application · medium

   A financial services firm fine-tunes a large language model for internal document summarization. The compliance team requires a complete record of who changed the model, what changed, and when. The firm uses Azure Machine Learning for model management. What should you configure to meet this requirement?

   • AEnable Azure Machine Learning's data labeling project for the training data.
   • BStore model files in a version-controlled Git repository without additional logging.
   • CEnable Azure Monitor for the compute cluster only.
   • DUse Azure Machine Learning's model registry with versioning and audit logs.
   Reveal Solution Discussion Select an answer first