OpenAI Adds Two Tiers to Cybersecurity Program

OpenAI has expanded its Daybreak cybersecurity research program with two new access tiers: Daybreak Blue and Daybreak Red. Blue targets defensive tasks like code scanning and malware analysis, while Red uses a specialized model, GPT-5.6-Cyber, for advanced tasks like zero-day vulnerability discovery. GPT-5.6-Cyber completed 95% of advanced cybersecurity benchmark requests, compared to 1.5% for the standard GPT-5.6 Sol. OpenAI says the model helped find two high-severity Chrome vulnerabilities and three critical database flaws. Enhanced security measures, including mandatory hardware key logins, roll out next month.