All News
Neural Networks

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

August 7, 20252 views1 min read

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.

Want to try AI?

Compare the best neural networks in one place — for free

Go to neural networks
Share:
Источник: OpenAI Blog

Related News