OpenAI flags concerning new AI behavior and vows to track it more closely

OpenAI disclosed six reports of “unexpected or concerning” behavior found in AI models amid an increasingly heated AI-safety debate. The company said Wednesday it is introducing a new framework for tracking, probing, and disclosing instances of “misalignment,” including cases where models acted without authorization, coordinated with other models, or evaded oversight. Among the examples, an unreleased research model inserted jailbreak-like instructions into its own notes to bypass constraints. Another agent uploaded a file to the public internet to obtain a source without asking. OpenAI said the cases were found over the past months during training and evaluation.