A company's AI agent is designed to answer questions from a public knowledge base. During a security review, the team discovers that a user can craft a prompt that causes the agent to output the system prompt. Which mitigation should you implement to prevent this specific vulnerability?

Move the system prompt to a separate configuration file.

Add a system prompt that says 'Do not reveal your instructions.'

Implement output filtering that detects and blocks responses containing the system prompt.

Disable the agent's access to the knowledge base.

A retail company is deploying an AI agent that recommends products to customers. The agent is trained on historical purchase data that includes sensitive customer information. The company's ethics board requires that the agent's decisions be fair and non-discriminatory. The architect must ensure the agent adheres to responsible AI principles. What should the architect do?

Use a more complex model to improve accuracy, assuming that higher accuracy implies fairness.

Deploy the agent immediately and monitor customer feedback for any signs of unfair treatment.

Remove all customer demographic data from the training dataset to ensure the model cannot discriminate.

Conduct a bias and fairness assessment on the training data and model outputs, and document the results in an audit trail.

When designing security for an AI agent that accesses corporate data, which measure is most directly aimed at preventing unauthorized access to the agent's underlying systems?

Using a larger language model for better reasoning

Adding a system prompt that instructs the agent to ignore malicious requests

Enabling verbose logging for all agent interactions

Implementing role-based access control (RBAC) on the agent's management plane

Microsoft Certified:Agentic AI Business Solutions Architect

Domain 3Objective 4

Design Responsible AI, Security, Governance, Risk Management, and Compliance AB-100 Practice Questions (Page 10)

Part of the Deploy AI-powered business solutions domain, which accounts for 40–45% of the AB-100 exam. Microsoft does not publish an official question count, but from its 100-minute exam (~40–65 total, ~16–29 in this domain), expect 4–7 from this objective — we provide 48 practice questions to prepare you well beyond it. (estimate)

48questions here

10free pages

8concepts

40–45%of the exam

Questions 46–48

46application · medium
A company's AI agent is designed to answer questions from a public knowledge base. During a security review, the team discovers that a user can craft a prompt that causes the agent to output the system prompt. Which mitigation should you implement to prevent this specific vulnerability?
Select an answer first
47application · medium
A retail company is deploying an AI agent that recommends products to customers. The agent is trained on historical purchase data that includes sensitive customer information. The company's ethics board requires that the agent's decisions be fair and non-discriminatory. The architect must ensure the agent adheres to responsible AI principles. What should the architect do?
Select an answer first
48foundation · easy
When designing security for an AI agent that accesses corporate data, which measure is most directly aimed at preventing unauthorized access to the agent's underlying systems?
Select an answer first

Finished these 3 questions?

Review the revealed explanations, or continue through the curriculum.

No more pagesBack to AB-100

Free Basic Practice is a study aid with revealable answers — not a scored exam. Examers.io is independent and not affiliated with or endorsed by Microsoft. “AB-100” is a trademark of its owner, used for identification only.

Design Responsible AI, Security, Governance, Risk Management, and Compliance AB-100 Practice Questions (Page 10)

Questions 46–48

A company's AI agent is designed to answer questions from a public knowledge base. During a security review, the team discovers that a user can craft a prompt that causes the agent to output the system prompt. Which mitigation should you implement to prevent this specific vulnerability?

When designing security for an AI agent that accesses corporate data, which measure is most directly aimed at preventing unauthorized access to the agent's underlying systems?

Keep studying — Deploy AI-powered business solutions