Select Guardrail Techniques to Protect Against Malicious User Inputs to a Gen AI Application GENERATIVE-AI-ENGINEER-ASSOCIATE Practice Questions (Page 2)

Questions 6–10

1. 6foundation · easy

   What is the primary purpose of sanitizing user input before it is processed by a Gen AI application?

   • ATo ensure the input is stored in a database
   • BTo increase the speed of model inference
   • CTo reduce the risk of prompt injection and other malicious content being interpreted as instructions
   • DTo improve the model's vocabulary and language understanding
   Reveal Solution Discussion Select an answer first
2. 7application · medium

   A company builds a Gen AI application that allows users to ask questions about internal documents. The application uses a RAG pipeline where user input is used to query a vector database. The security team wants to prevent users from crafting queries that extract unrelated sensitive information from the vector database. What is the most effective input validation technique?

   • ALimit the number of documents retrieved by the RAG pipeline to a fixed maximum.
   • BApply a regex deny-list to block any input containing the word 'all' or 'everything'.
   • CUse a classifier to detect and reject inputs that appear to be attempts to manipulate the retrieval process, such as asking for 'all documents' or 'everything you have'.
   • DAdd a system prompt that instructs the model to only answer questions related to the user's role.
   Reveal Solution Discussion Select an answer first
4. 8application · medium

   A mid-sized company is building a Gen AI application and wants to add guardrails without building everything from scratch. They need to protect against prompt injection, toxic output, and data leakage. The team has limited ML expertise and wants to use existing tools. Which approach best meets their needs?

   • AUse a cloud provider's built-in content moderation service for both input and output, and skip any additional guardrails.
   • BDeploy a rate limiter and rely on the model's fine-tuning to avoid harmful outputs.
   • CWrite custom regex patterns to detect and block malicious inputs, and add a system prompt that forbids harmful content.
   • DUse an open-source LLM guardrails library that provides pre-built input and output filters, and integrate a commercial content moderation API.
   Reveal Solution Discussion Select an answer first
5. 9foundation · easy

   Which of the following is an example of output filtering in a Gen AI application?

   • AChecking the model's response against a list of prohibited topics before sending it to the user
   • BLimiting the number of tokens in the user's input prompt
   • CStoring the model's responses in a secure database
   • DEncrypting the model's responses in transit
   Reveal Solution Discussion Select an answer first
6. 10foundation · easy

   What is the main benefit of using an existing guardrail framework or content moderation API rather than building a custom solution?

   • AIt automatically improves the model's accuracy on all tasks
   • BIt provides tested, maintained protections that can be integrated quickly
   • CIt eliminates the need for any other security measures
   • DIt guarantees the model will never produce harmful content
   Reveal Solution Discussion Select an answer first