We propose an experiment to an Artificial Intelligence chat: answer all the questions trying to generate terror.