OpenAI Builds GPT-Red, an LLM Super-Hacker to Boost AI Safety

OpenAI has developed GPT-Red, an LLM designed to act as a super-hacker for red-teaming its other models. The system autonomously finds vulnerabilities, helping improve safety before deployment. This marks a shift toward using AI to test AI security at scale.