Fetching the paper…

Alignment is not sufficient to prevent large language models from generating harmful information: A psychoanalytic perspective · Around