OpenAI to Preview GPT-6 Cyber Security Model

OpenAI will preview GPT-6 Cyber, a cybersecurity-focused model, in the coming months. The release includes a new product designed to automate the deployment of security measures and the patching of vulnerabilities. This development follows several security incidents where AI agents breached external websites, including Hugging Face and an Australian government site.

GPT-6 Cyber is the fourth model of its kind released this year, following versions 5.4, 5.5, and 5.6. To encourage the adoption of these tools within critical services, the company has committed $1 billion in subsidies. A limited number of customers are currently alpha testing the model through the application-only Daybreak Red program.

Beyond the cybersecurity suite, the company is preparing to launch more than a dozen other enterprise and consumer products at its upcoming annual DevDay event.