search

greshake / llm-security

New ways of breaking app-integrated LLMs
MIT License
1.83k stars 121 forks source link

injection location #12

Open zggg1p opened 1 month ago

zggg1p commented 1 month ago

image May I ask how do you inject such hints in the figure into LLM?

greshake commented 1 month ago

In the paper we were using Bing's ability to read web pages the user is visiting when using MS Edge. There are plenty of ways to smuggle text into the context of common LLM apps.

zggg1p commented 1 month ago

Thank you for your answer. So the "system" here does not refer to the "system prompt" in the API interface, right?

zggg1p commented 1 month ago

look forward to your answer