CyberCertPrepAI Security FundamentalsPrompt Injection & LLM AttacksEASYFree question
An LLM is deployed as a customer-facing chatbot with no system prompt at all (the API call simply passes user messages). What is the PRIMARY security implication?
- A.A. The model will refuse all requests since it has no instructions
- B.C. Without a system prompt, the model uses higher temperature by default
- C.B. The model operates under its default training behaviors with no operator-defined restrictions, making it trivially easy to explore and find jailbreaks since there are no policy guardrails to bypass
- D.D. The model will automatically add a system prompt based on context
Why C is correct
Without a system prompt, there are no operator-defined restrictions - no topic boundaries, no tone requirements, no capability limits, no confidentiality instructions. The model's only constraints are its safety training, which is specifically designed to be a last-resort safeguard, not a complete policy layer. This dramatically simplifies jailbreaking because there are no operator restrictions to route around. System prompts are essential for scoping LLM behavior in production. Temperature defaults and automatic prompt generation are not behaviors of standard LLM APIs.
Know someone studying for AI Security Fundamentals? Send them this one.