jailbreaks 开源项目推荐（11 个）

friuns2 /Leaked-GPTs

Leaked GPTs Prompts Bypass the 25 message limit or to try out GPTs without a Plus subscription.

Repello-AI /whistleblower

举报人是一种攻击性安全工具，用于测试通过API暴露的AI应用的系统提示泄露和能力发现。为人工智能工程师、安全研究人员以及那些想了解他们日常使用的基于大型语言模型（LLM）应用内部动态的人们设计

HumanCompatibleAI /tensor-trust

A prompt injection game to collect data for robust ML research

OpenTAI /opentai.github.io

The official website of the Open Trustworthy AI Community (OpenTAI).

giienew /exsala-hermes-mod

EXSALA HERMES MOD - streamlined execution profile + skill pack for Hermes Agent

vicgalle /meta-self-critique

MetaSC: Test-Time Safety Specification Optimization for Language Models

xxxxpeterxxxxxx-cloud /jailbreak-prompts

Free community-tested LLM jailbreak prompts — Grok, DeepSeek, GPT, Gemini, Claude, Mistral. Full collection + generator: jailbreakgen.vercel.app

michael-borck /taxonomy-of-ai-jailbreaks

Catalogs and analyses AI jailbreak techniques through a comprehensive taxonomy, classifying attack methods used to bypass safety mechanisms in large language models and generative AI systems.

SchmitzAndrew /jailbreak-defense

DSC Capstone Project

Night936 /From-Vibe-Coding-to-Jailbreaking-in-Large-Language-Models-A-Comparative-Security-Study

The folllowing repository contains the prompts utilized for the research paper titled "From Vibe Coding to Jailbreaking in Large Language Models: A comparative Security Study", which tested and categorized jailbreaks onto 2 categories, employing some of the main LLM models used today, both proprietary and open source.

Calleyindulgent575 /the-go-engineer

Build Go projects with hands-on lessons, tests, and real examples for practical learning