Question 5
Domain 2 — AI Operations, Lifecycle, and Control EnvironmentPrompt injection attacks are a security concern primarily for which type of AI system?
Correct answer: C
Explanation
Prompt injection attacks target systems that follow natural-language instructions, especially models that can be steered by user input. Large language models and generative AI systems are most exposed because they process prompts and may be manipulated into ignoring prior instructions or revealing hidden data.
Why each option is right or wrong
A. Traditional machine learning classification models
B. Computer vision models processing images
C. Large language models (LLMs) and generative AI systems
Prompt injection is a known attack vector against instruction-following models that accept free-form natural-language input and then condition their outputs on that input, which is characteristic of large language models and other generative AI systems. In practice, the attacker’s prompt can override or subvert the system prompt, causing the model to disclose hidden context, execute unintended actions, or ignore prior constraints—risks that are far less relevant to deterministic, non-generative systems.
D. Reinforcement learning agents in robotics