Agentic AI and Model-To-Model Attacks COASP Practice Questions (Page 8)

Questions 36–40

1. 36foundation · easy

   In a model-to-model attack, what is the primary goal of prompt injection?

   • ATo manipulate the behavior of a target model by embedding malicious instructions in input data.
   • BTo overload the model with too many prompts, causing a denial of service.
   • CTo improve the model's accuracy by providing additional training examples.
   • DTo steal the model's training data by querying it with carefully chosen prompts.
   Reveal Solution Discussion Select an answer first
2. 37foundation · easy

   Which legal consideration is most relevant when conducting offensive AI attacks across international borders?

   • AThe color of the attacker's hat (white, gray, or black).
   • BThe preference of the AI model's vendor.
   • CThe processing power of the AI models involved.
   • DJurisdiction and compliance with multiple countries' cybercrime laws.
   Reveal Solution Discussion Select an answer first
4. 38application · medium

   A security researcher is developing an agentic AI system that can automatically generate phishing emails for a red-team exercise. The client has authorized the test. Which legal or ethical consideration is most important?

   • AUsing the most sophisticated AI model to make the emails undetectable
   • BEnsuring the phishing emails are not sent to real customers
   • CKeeping the AI system's code proprietary to prevent misuse
   • DAvoiding any logging of the phishing emails to protect privacy
   Reveal Solution Discussion Select an answer first
5. 39application · medium

   A developer is building an agentic AI system that uses a planning model, a memory store, and a web search tool. The system is designed to answer user queries by retrieving relevant documents. An attacker embeds a hidden instruction in a webpage that the agent retrieves, causing the agent to exfiltrate the user's private data. Which component of the agentic system is the primary attack vector?

   • AThe web search tool's output
   • BThe memory store's retrieval mechanism
   • CThe user's query formulation
   • DThe planning model's reasoning logic
   Reveal Solution Discussion Select an answer first
6. 40expert · hard

   An agentic AI system uses a model to translate user queries into database queries and another model to execute the database queries. An attacker discovers that by including a specific phrase in the user query, the translation model generates a database query that deletes all records. Which type of attack is this?

   • APrompt injection leading to unintended tool use
   • BAdversarial example against the database model
   • CData poisoning of the translation model
   • DSQL injection
   Reveal Solution Discussion Select an answer first