Lessons From Red-teaming Auto-review, OpenAI's Blocking Monitor

OpenAI conducted red-teaming exercises on its Auto-review and Blocking Monitor systems to identify vulnerabilities and improve AI content moderation. These efforts aim to enhance the robustness and safety of automated content filtering.