AI.RM
AI Ready Media
From Hard Refusals to Safe-Completions: Output-Centric Safety Training — OpenAI • Libertify