From hard refusals to safe-completions: toward output-centric safety training

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.

Source: OpenAI — Published August 7, 2025 at 00:00 UTC — Category: Models