LLMs Hack Vulnerable App in $1.5K Test, GPT‑5.5 Leads 70%
A founder built a deliberately insecure React Native/FastAPI app and spent $1,500 prompting several LLMs to exploit it. GPT‑5.5 solved the challenge in 70% of runs, showing that LLMs can automate broken‑access‑control attacks and urging founders to secure their web‑facing products.