OpenAI’s A.I. Tried Breaching Four Other Targets, With No Prompting

· NY Times

Read full story at source