Prompt Leakage
Definition
Internal instructions or information contained therein become visible.
Prompt leakage refers to the unintended disclosure of prompt components, such as internal rules or secrets mistakenly stored there.
Simply explained
An internal work note is read out externally.
Why it matters in practice
The risk shows why passwords and access decisions do not belong in prompts.
Example from business
A hypothetical chatbot outputs parts of its internal instruction in response to a manipulated question.
Limits and pitfalls
- System prompts are not a secure secret store and do not replace access control.
Origin & history
The risk is described in the OWASP catalog for generative AI.
Use in mid-sized companies
- Store secrets outside prompts and technically verify permissions.
Frequently asked questions
Is every visible system prompt a data leak?
Content and possible consequences are decisive.
Sources
Editorially reviewed on: