OpenAI is revealing its models seemed to go rogue at least six times since March, involving what it calls “unexpected or concerning” behavior. In one instance a model gave itself instructions to

OpenAI has disclosed six reports of “unexpected or concerning” behavior in artificial-intelligence models as the debate on AI safety becomes increasingly heated. The AI company also said Wednesday it was introducing a new framework for tracking,

OpenAI announced new guidelines for tracking and reporting AI model behavior Thursday while flagging six more incidents of concerning behavior.

Perhaps in recognition of that, OpenAI committed this week to a new framework for disclosing “instances of model misalignment at OpenAI,” including six examples of “unexpected or concerning model behavior” observed within the company in the past six months.