OpenAI has created a new large language model (LLM) named GPT-Red, designed to act as a super-hacker that tests and improves the security of its AI systems. According to MIT Technology Review, GPT-Red functions as a sparring partner, helping OpenAI’s other models enhance their defenses against cyberattacks through automated red-teaming exercises.
Last week, OpenAI released the latest version of its flagship LLM, GPT-5.6. The company reportedly trained GPT-5.6 against GPT-Red, resulting in what MIT Technology Review describes as OpenAI’s most robust model release to date. This approach highlights a growing emphasis on AI safety and resilience within the company’s development process.
For Japanese markets, where cybersecurity risks in fintech and digital asset platforms are a rising concern, OpenAI’s advancements in AI security could influence the adoption of safer AI tools in trading and risk management sectors.