Security & Privacy
Prompt leakage
When hidden instructions or system prompts are unintentionally revealed.
Example
A chatbot accidentally includes its confidential system instructions in a response.
Why it matters
It helps teams recognize unintended disclosure of a product's hidden guidance.
You'll hear it
“The reply included instructions that were meant to stay internal.”
Can you control it?
Set by the team
The people building or running the AI choose this setup. An everyday user generally needs their help to change how this part works.
Common questions
- Does a leaked prompt reveal the entire model?
- No. Instructions are separate from the model's learned internal values.
- Can an AI invent supposed hidden instructions?
- Yes. A claimed disclosure needs verification; confident wording does not prove it is genuine.
- Does hiding instructions guarantee the behavior they describe?
- No. Secrecy and reliable instruction-following are different issues.