A one-prompt attack that breaks LLM safety alignment

(microsoft.com) by weinzierl | Feb 10, 2026 | 0 comments on HN