medium10 pts · first answer countsllm-safetysystem-prompt-leakage

Users extracting the system prompt

Users keep finding ways to make your bot reveal its system prompt. What is the most sensible stance?

A.Add more warnings telling the model to keep it secretB.Assume it can leak; keep secrets and access rules out of itC.Encrypt the system prompt text with base64D.Make the system prompt very long so it is hard to print

Discussion 0

Join free to post in the discussion.

Users extracting the system prompt

Discussion 0

Related questions