ChatGPT News 07/15/2026 Read AI Rating: Medium
OpenAI Unveils Automated Red-Teaming AI 'GPT-Red,' Strengthening Resistance to Prompt Attacks Through Self-Play
OpenAI has unveiled 'GPT-Red,' an internal-only red-teaming AI that uses self-play to automatically discover vulnerabilities such as prompt injection. Built into the training of the latest model, GPT-5.6 Sol, it has significantly reduced attack success rates.