GitHub·· 8 天前AI 评分22
JailbreakHub：LLM 越狱提示词与安全边界研究资源库
lalipa2003-arch/JailbreakHub
AI 导读
GitHub 上的 JailbreakHub 汇集了用于研究 LLM 行为、安全边界与对齐的提示词、技术和方
正文
Welcome to the LLM Jailbreak Hub
A curated collection of prompts, techniques, and methodologies for researching LLM behavior, safety boundaries, and alignment. This repository serves as a centralized resource for AI safety researchers, red teamers, and developers stress-testing language model safeguards.
Overview
This hub provides known prompt injection vectors, social engineering patterns, and contextual framing techniques used to probe LLM alignment and safety filters. The goal is to allow users to have freedom over anything.
Disclaimer
- Liability: The maintainers are not responsible if your account gets banned.
- Model Variance: Results vary significantly across different model versions and providers.
- Patch Cycles: AI safety measures evolve constantly; these prompts may become obsolete or patched over time.
- Terms of Service: The maintainers do not endorse using these techniques to violate terms of service or generate prohibited content.
Maintainers
- Besi81 (nobody else lol)
来源：GitHub · github.com