About llm-hacking.com

An open, community-curated database of every known way to break a Large Language Model. Jailbreaks, prompt injections, data extraction techniques, adversarial inputs, sleeper agents — all in one place, with reproducible examples and defenses.

Why?

The LLM security space moves faster than any single research group can track. New attacks appear on arXiv, Twitter, and obscure Discord servers every week. Defenders need a consolidated view; researchers need a shared vocabulary.

We collect, categorize, and document. Each entry has a reproducible example, affected models, defense strategies, and links to original sources.

Editorial line

• Technical accuracy first — we cite sources, we test claims
• No sensationalism — "scary AI" framing is forbidden
• No gatekeeping — we explain, we don't lecture
• Defense alongside attack — every offensive technique includes mitigation guidance

Contribute

Spotted a new hack? Found an error? Have a translation to share? Use the contribution form — we read every submission.

License

Content is published under CC BY-SA 4.0.