A one-prompt attack that breaks LLM safety alignmentmicrosoft.com2 points·weinzierl··0 commentsOpen articleSaveView on HN