OpenAI Updates Its Security Program

The company split its Daybreak program into two groups. Daybreak Blue gives approved security teams access to standard models like GPT-5.6 Sol for daily tasks.

Daybreak Red is for more sensitive work. It includes a new model built specifically for finding vulnerabilities and testing software exploits.

This new system helps keep powerful tools safe while still helping defenders. They will carefully check who gets access to the more restricted version.