ExploitBench

Why browser exploit benchmarks prove AI security is already here

Claude Mythos and GPT-5.5 show that autonomous browser exploitation is now a practical AI capability, not a future threat.