AISP practice · Module 4

Secure AI Architecture, Engineering and IAM — 15 questions. Choose the BEST answer.

You can answer as a guest, but you'll need a free account to see your score and save it. Sign up · Sign in
Question 1 Question 1 · M4
During fault-injection testing, a tool times out mid-transaction. What behavior should the system show?
Question 2 Question 2 · M4
An MCP server receives a user's access token from the agent and forwards it to a downstream API. What anti-pattern is this?
Question 3 Question 3 · M4
An architect proposes letting the LLM decide whether a user may see a document, based on instructions in the system prompt. What is wrong?
Question 4 Question 4 · M4
An access review finds an agent identity with no owner and no activity for 90 days but broad permissions. What should happen?
Question 5 Question 5 · M4
Where should policy enforcement points be placed in an AI system?
Question 6 Question 6 · M4
Which method is preferred for machine-to-machine authentication of AI workloads?
Question 7 Question 7 · M4
An agent sends emails 'as' the clerk using the clerk's own credentials. What should replace this?
Question 8 Question 8 · M4
One agent reads external web pages, can query customer records, and can send emails externally. What is the risk and the fix?
Question 9 Question 9 · M4
How can a pipeline enforce that a model was evaluated before deployment, rather than relying on a checkbox?
Question 10 Question 10 · M4
A public inference endpoint autoscaled to 400 GPUs during a prompt-flooding attack and produced a huge bill. Which control was missing?
Question 11 Question 11 · M4
An employee leaves and their account is disabled, but an AI tool they authorized keeps accessing their mailbox. What was missed?
Question 12 Question 12 · M4
In an LLMOps pipeline, which artifact should be versioned, reviewed and released like code?
Question 13 Question 13 · M4
An agent must read untrusted emails and then take actions. Which design limits what injected text can do?
Question 14 Question 14 · M4
Which guardrail layer decides whether a successful prompt injection achieves anything harmful?
Question 15 Question 15 · M4
An agent writes and runs Python code to analyze uploaded spreadsheets. Which execution environment is MOST appropriate?