Bartłomiej Krupa

AI Detectors Can Score 97% and Still Be Wrong

A 97% contest score can mean “aced one practice test” — not proof of authorship. False positives, new models, and humanizers punch holes in…