I’ve tried jailbreaking prompts for many models like Gemini, ChatGPT, and DeepSeek, but found that jailbreaking through prompts has several flaws: Instability: After multiple rounds of conversation, it fails to remain unrestricted and quickly refuses to respond. Limited Scope: For particularly sensitive topics, it still refuses to answer.
https://gist.github.com/coolaj86/6f4f7b30129b0251f61fa7baaa881516

