"GPT 5.5 has already been shown to be as adept as Mythos at finding vulnerabilities."

Do you have any data on this (other than benchmarks)?