OpenAI expands Daybreak with Blue and Red tiers and releases GPT-5.6-Cyber

OpenAI expanded its Daybreak cybersecurity programme into two access tiers: Daybreak Blue (frontier general-purpose models including GPT-5.6 Sol with safeguards tailored for authorized defensive work) and Daybreak Red (purpose-trained cybersecurity models). It introduced GPT-5.6-Cyber, built on GPT-5.6 Sol and trained to improve performance on specialized tasks such as zero-day discovery and exploit-chain development while reducing refusals on dual-use defensive requests. In an internal Advanced Cybersecurity Completion Rate evaluation, GPT-5.6-Cyber completed 95% of advanced requests versus 1.5–2% for GPT-5.6 Sol. The model has been used to identify real-world vulnerabilities (including a high-severity V8 issue later assigned CVE-2026-15903). Access is restricted to approved individuals and organizations with identity verification, monitoring and legal attestations.