OpenAI Builds GPT-Red to Automate Red-Teaming

OpenAI built GPT-Red, an automated red-teaming model that scales prompt injection vulnerability discovery. GPT-Red adversarially trained GPT-5.6 Sol, making it the most robust model to prompt injections to date, achieving 6x fewer failures against direct prompt injection benchmark compared to GPT-5.5. GPT-Red is kept separate from other models to prevent malicious capabilities from reaching bad actors.