QuestionQ56

VerifiedSecurity, Compliance, and Governance for AI Solutions

Which prompting attack directly exposes the configured behavior of a large language model (LLM)?

Explanation

Extracting the prompt template is a prompting attack that directly reveals the underlying configured behavior, instructions, or system prompts of an LLM. This exposes how the model has been configured and designed to respond, giving attackers insight into the model's constraints and behavioral instructions. The other options involve manipulating the model's behavior (persona switches, friendliness exploitation) or ignoring prompts, but they do not directly expose the configured behavior itself.

Learn more

Community Discussion

No comments yet. Be the first to start the discussion!