When AI Grades AI, the Safety Problems Disappear
100 mental health professionals graded LLM answers to real patient questions. AI judges systematically overrated them and missed safety problems experts caught.
Matthew Sexton, LCSWRead article
AI evaluation
Still neon
1 article, newest publications first. Browse this subject or search the full archive.
4 articles in the wider archive have updated wording and source notes. The other 164 have not been fact-checked as part of this review. Older topics and links are retained to keep existing articles findable.