Select Guardrail Techniques to Protect Against Malicious User Inputs to a Gen AI Application GENERATIVE-AI-ENGINEER-ASSOCIATE Practice Questions (Page 1)

Questions 1–5

1. 1application · medium

   A social media company deploys a Gen AI assistant that generates user-facing responses. The assistant is powered by a fine-tuned model that occasionally produces toxic or offensive content when users input adversarial prompts. The team wants to add a guardrail that specifically prevents harmful content from reaching users. Which approach should they choose?

   • AAdd a system prompt that instructs the model to always be polite and respectful.
   • BAdd a post-processing step that uses a content moderation API to score the model's output and block or replace any response that exceeds a toxicity threshold.
   • CUse input validation to reject any input that contains offensive words.
   • DImplement a rate limiter that throttles users who submit multiple requests in a short time.
   Reveal Solution Discussion Select an answer first
2. 2foundation · easy

   Why is it important to treat user input as untrusted data in a Gen AI application's prompt design?

   • ABecause user input is always encrypted and cannot be read
   • BBecause user input is always stored in a database
   • CBecause user input may contain instructions that attempt to override the system prompt
   • DBecause user input is always too long to process
   Reveal Solution Discussion Select an answer first
4. 3application · medium

   An e-commerce company deploys a Gen AI chatbot that helps customers with product recommendations. The chatbot sometimes generates responses that include offensive language when users input adversarial prompts. The company wants to ensure that no offensive content is ever shown to customers. Which guardrail technique is most effective for this requirement?

   • AUse input validation to reject any input that contains offensive words.
   • BUse a content moderation API to filter the model's output and block any response that contains offensive language.
   • CAdd a system prompt that instructs the model to never use offensive language.
   • DImplement a rate limiter that blocks users who send multiple requests in a short time.
   Reveal Solution Discussion Select an answer first
5. 4foundation · easy

   Which of the following is an example of a guardrail tool or framework used to protect a Gen AI application?

   • AA database used to store user prompts
   • BA content moderation API that flags toxic or unsafe text
   • CA load balancer that distributes traffic across servers
   • DA monitoring dashboard that shows model latency
   Reveal Solution Discussion Select an answer first
6. 5expert · hard

   A healthcare company is deploying a Gen AI application that assists doctors with clinical documentation. The application accepts free-text input and uses a RAG pipeline to retrieve relevant medical literature. The company must comply with strict data privacy regulations. They are concerned about both prompt injection attacks and the model inadvertently generating responses that contain protected health information (PHI). Which combination of guardrails is most effective?

   • AImplement a role-based access control (RBAC) layer on the retrieval system to restrict which documents each user can access, and add an output filter that redacts any PHI from the model's response.
   • BUse a content moderation API on the input and output, and log all queries for audit.
   • CUse a deny-list of PHI-related terms and a system prompt that instructs the model to avoid PHI.
   • DImplement a rate limiter and a CAPTCHA to prevent automated attacks.
   Reveal Solution Discussion Select an answer first