OpenAI ↗ · 2025-08-07 08:00 · UTC+8
From hard refusals to safe-completions: toward output-centric safety training
Company news OpenAI Source language: English
Discover how OpenAI's new safe-completions approach in GPT-5 improves both safety and helpfulness in AI responses—moving beyond hard refusals to nuanced, output-centric safety training for handling dual-use prompts.
Excerpt
Discover how OpenAI's new safe-completions approach in GPT-5 improves both safety and helpfulness in AI responses—moving beyond hard refusals to nuanced, output-centric safety training for handling dual-use prompts.
Short excerpt from the source — open the original for the full text.