3 ms·Llama Guard: LLM-Based Input-Output Safeguard for Human-AI Conversations2 points by georgehill 3y ago