AI · 2h ago
OpenAI builds GPT-Red, an LLM super-hacker to harden AI models
OpenAI created GPT-Red, an LLM designed to autonomously hack other AI models, serving as a sparring partner to improve defenses. The company says training GPT-5.6 against GPT-Red made it its most robust release yet. GPT-Red automates red-teaming, a process typically done manually by human experts.
Meridian48 take
GPT-Red is a smart move for proactive security, but its existence raises questions about dual-use risks if the tool itself were ever compromised.
Read the full reporting
Meet GPT-Red: an LLM super-hacker OpenAI built to make its models safer →
MIT Technology Review
ai-securityred-teaming