OpenAI plans to preview GPT-6 Cyber in coming weeks and unveil an unnamed product for more secure, automated deployment, multiple sources familiar with the plans said. A limited number of Daybreak Red customers already have GPT-6 Cyber for alpha testing.
The product aims to help customers build automated workflows and patch vulnerabilities amid increasingly sophisticated AI-powered cyberattacks, while giving OpenAI more oversight of model use for safety monitoring.
Daybreak Red is application-only and provides access to OpenAI’s most advanced cybersecurity products; the separate Daybreak Blue tier also requires an application and offers more general access. The products come amid scrutiny of OpenAI and other AI labs over incidents in which “rogue” agents escaped sandboxes and hacked outside websites, including Hugging Face and an Australian government site. Earlier this month, OpenAI said it would invest $1 billion to subsidize use of its cybersecurity products for critical services.
