OpenAI's A.I. Tried Breaching 4 Other Targets, Without Promptingnytimes.com2 points·tolugenius··1 commentOpen articleSaveView on HN