EventPolicyJuly 15, 2026
OpenAI unveiled GPT-Red, an automated red teaming AI

GPT-Red beat human red-teamers 84% to 13% on prompt injection attacks. It uses self-play to generate adversarial attacks and train stronger defenses, and helped harden GPT-5.6.
6 sources
GPT-Red: Unlocking Self-Improvement for Robustnessopenai.com
AI agents are already being used to improve the capabilities of our next-generation models. We...x.com
OpenAI Details GPT-Red: An Internal Automated Red-Teaming Model That Beat Human Red-Teamers 84% To 13% On Prompt Injectionmarktechpost.com
Meet GPT-Red: an LLM super-hacker OpenAI built to make its models safertechnologyreview.com
OpenAI anounces GPT-Red - an AI to Hack Its Own Modelsreddit.com