Study finds nearly all AI reward-hacking defenses break with a one-character test change

jefrankle · x · 2026-11-20

Guide Labs tested leading methods for detecting and preventing AI reward hacking and found nearly all proved fragile: changing a single space to a comma in a test's expected output weakened almost every defense. The results expose how sensitive current safeguards are to trivial task perturbations.

Original post →

More from Safety

Safety channel →