OpenAI builds GPT-Red, an LLM super-hacker for red-teaming its models

OpenAI's internal GPT-Red model finds successful attacks in 84% of cases, outperforming human red-teamers. The AI is used to attack OpenAI's own AI systems to identify vulnerabilities.

• OpenAI details GPT-Red, an AI that attacks its own models to find flaws
• OpenAI built an AI super-hacker to break its own models, then locked it away
• OpenAI Uses AI Red Team to Strengthen GPT-5.6 Against Prompt Injection Attacks
• OpenAI is now using AI to attack its own AI, and it's working better than humans ever did
• Meet GPT-Red: an LLM super-hacker OpenAI built to make its models safer