Design Responsible AI, Security, Governance, Risk Management, and Compliance AB-100 Practice Questions (Page 2)

Questions 6–10

1. 6foundation · easy

   When deploying an AI solution that processes personal data of EU citizens, which action is most aligned with data residency compliance?

   • AAnonymizing the data before processing
   • BSelecting a cloud region within the EU to store and process the data
   • CUsing a model hosted in the US but with data stored in the EU
   • DEncrypting the data with a strong algorithm
   Reveal Solution Discussion Select an answer first
2. 7expert · hard

   A multinational bank is deploying an AI agent that processes loan applications. The bank operates in the EU and Brazil. The agent uses a large language model hosted in Azure OpenAI in the East US region. The bank's compliance team requires that personal data of EU customers remain within the EU and that all model decisions be explainable. The architect must design a solution that meets these requirements while minimizing latency for EU users. What should the architect do?

   • ADeploy the agent and its grounding data in an Azure region within the EU, use Azure OpenAI in the same EU region, and enable detailed logging of all model inputs and outputs for audit purposes.
   • BUse a Brazilian region for the Azure OpenAI resource because Brazil is closer to the EU than the US, and rely on the bank's existing EU data center for storage.
   • CUse a custom model trained on-premises in the EU and deploy it via Azure Stack Hub, but do not log any prompts to avoid storing personal data.
   • DKeep the Azure OpenAI resource in East US but use Azure Front Door to route EU traffic to a proxy in the EU that logs all requests and responses.
   Reveal Solution Discussion Select an answer first
4. 8expert · hard

   A healthcare provider uses an AI agent to triage patient symptoms. The agent sometimes gives advice that conflicts with clinical guidelines. The provider wants to ensure the agent operates within ethical and organizational boundaries. What governance mechanism should you implement?

   • ALet clinicians manually override the agent's advice without logging.
   • BAllow the agent to give any advice but include a disclaimer.
   • CDefine a policy that the agent must only provide information that aligns with approved clinical guidelines, and implement a review process for any deviations.
   • DDisable the agent for all patients.
   Reveal Solution Discussion Select an answer first
5. 9application · medium

   A customer service company is deploying an AI agent that accesses customer order history to answer questions. The agent is integrated with a backend database. The security team wants to ensure that the agent cannot be manipulated to perform unauthorized actions, such as deleting orders. What should the architect implement?

   • ADisable the agent's access to the database and have it rely on cached data that is refreshed daily.
   • BUse a system prompt that instructs the agent to never delete or modify data.
   • CGrant the agent read-only permissions on the database and use Azure API Management to validate and filter incoming requests.
   • DGive the agent full database permissions so it can handle all customer requests efficiently.
   Reveal Solution Discussion Select an answer first
6. 10application · medium

   A cybersecurity firm is deploying an AI model that detects malware. The model is hosted in Azure and accessed by external customers via an API. The firm wants to prevent attackers from reverse-engineering the model by sending carefully crafted inputs. What should the architect implement?

   • AAllow unlimited API calls without authentication to encourage usage.
   • BUse a smaller model that is less accurate but harder to reverse-engineer.
   • CProvide the model's confidence scores to all API users to increase transparency.
   • DImplement rate limiting, require API keys, and monitor for anomalous request patterns that may indicate model extraction attempts.
   Reveal Solution Discussion Select an answer first