Meet GPT-Red: an LLM super-hacker OpenAI built to make its models safer

OpenAI built an LLM super-hacker called GPT-Red to boost its models' defenses against cyberattacks. The company released GPT-5.6 last week, its most robust release yet, trained against GPT-Red. GPT-Red automates the process of finding vulnerabilities, helping OpenAI create safer AI systems.