AI Risk Benchmark: GPT-5 Leads, but Misalignments Persistsubstack.com4 points·m1chael3ma··0 commentsOpen articleSaveView on HN