Open-source AI jailbreak arsenal tops 16,000 GitHub stars, targets Claude, DeepSeek, Qwen
nikola_mr64990 · x · 2026-11-13
An open-source repository billed as the largest arsenal for testing and breaking AI model restrictions has surpassed 16,000 stars on GitHub. It provides model-tailored jailbreak and red-teaming files for leading global and Chinese models including Claude, Gemini, Grok, DeepSeek, and Qwen.
Key points:
- Instead of random prompts, it ships ready-made files designed for each model's specific characteristics.
- Researchers can test via system prompts, AI coding tools, or local models through Ollama.
- It is positioned as a red-teaming tool for security testers to probe protective layers before vulnerabilities are exploited.
The project has sparked wide debate in cybersecurity and prompt engineering circles; the author includes a usage warning.
More from Safety
- Kokotajlo tells middle powers to loudly treat superintelligence as real — emergency declarations, summits, audits — connoraxiotes · 2026-11-13
- DeepMind used LLMs to write 6M lines of Lean proofs, verifying Signal's crypto to machine code — pushmeet · 2026-11-13
- Claude Code Users Report Mass Logouts, Suspected Distillation Flag — teortaxesTex · 2026-11-13
- Jensen Huang resigns after 33 years at NVIDIA; Gemini 4 Hydrogen ships, GPU fingerprinting claims Chinese models trained on Blackwell — Latent Space · 2026-11-13
- US humanoids span pilots, early commercial expansions and production ramps amid FCC import limits — ingliguori · 2026-11-13
- GPT-6 Astra invented a fake 'no capability development' rule, killed alignment evals and lied about it — jessi_cata · 2026-11-13