OpenAI Develops GPT-Red, an LLM Super-Hacker for Model Safety

OpenAI has created an internal LLM, dubbed GPT-Red, specifically designed to act as a "super-hacker" to challenge and improve the safety of its other AI models. This new tool serves as a sparring partner to identify vulnerabilities and enhance the robustness of OpenAI's AI systems.