LLM Jailbreak Attacks

We propose AutoDefense, a response-filtering based multi-agent defense framework that filters harmful responses from LLMs.