OpenAI Develops AI Model for Red-Teaming Purposes
OpenAI has developed a new model called GPT-Red, designed to test and improve the robustness of other AI models by identifying vulnerabilities through red-teaming. The company used this approach on its own GPT-5.6 Sol model, enhancing its resistance against prompt injections.
Topics: AI GPT-Red OpenAI robustness testing