4 ms·

Can’t you just ask another LLM to analyse the text of the input to determine if it’s an attempted prompt injection?

• robterrell 3y ago
  That's a possible mitigation mentioned in the article.
  • robinduckett 3y ago
    Maybe I'll read it next time :D