We break AI before it breaks you.

Independent adversarial research on frontier models. Daily vulnerability disclosures, red-team teardowns, and a community of 2,000+ AI security researchers.

Cripple.AI is an independent collective publishing CVE disclosures on LLM jailbreaks, prompt injection, tool abuse, supply-chain compromises, and model-theft vulnerabilities across OpenAI, Anthropic, Google, Meta, Microsoft, Vercel, and the broader frontier-AI stack.