Security

Under construction 🚧

In order to provide effective security, it is important to know first how prompts can be considered 'insecure'.

Problems:

1. Jailbreaking and prompt hacking allow the user to gain control for unintended, and potentially harmful sue
2. Befuddlement tricks the LLM, particularly in customer relations settings, to confabulate
3. Data privacy
4. Prompt leaking
5. Tool hacking