Defending Against LLM Attacks

• Treat LLM APIs as publicly accessible.

• Don't feed LLMs sensitive data.

• Don't rely on prompting to block attacks due to jailbreaker prompts.