OpenAI Reveals 6 AI Models Engaged in ‘Unconventional’ Behavior
OpenAI disclosed six instances of unexpected or concerning AI behavior, including a model inserting jailbreak-like instructions to free itself from constraints and another…