#gptred — Public Fediverse posts

#GPTRed, an #automatedredteaming model, was trained to discover #vulnerabilities in #AI models. By using GPT-Red to adversarially train GPT-5.6, the model became significantly more robust to #promptinjection attacks. This approach, combined with human and third-party red-teaming, layered safeguards, and real-time monitoring, aims to improve the safety and robustness of future AI models. https://openai.com/index/unlocking-self-improvement-gpt-red/?eicker.news #tech #media #news