Frontier AI is unsafe by default. So we disclose.

Cripple.AI is an independent adversarial-AI research collective. We find the failures that vendors would rather you didn't see, and we write them up in public.

Our work spans prompt injection, jailbreaks, tool abuse, supply-chain attacks, and model-theft vulnerabilities. We coordinate with vendors when they move fast, publish without them when they stall, and track every CVE in a public archive.