Your question is LLM Prompt Injection Defenses. Take a moment with it on the right.
Talk me through your thinking if you like. When you're confident, submit your answer and I'll grade it like a real screen (7/10 or better passes).
What security controls would you implement to prevent prompt injection and data exfiltration in features utilizing large language models (LLMs) or artificial intelligence at Tools for Humanity?