OpenAI launches GPT-5.6-Cyber cybersecurity model with tiered access

OpenAI released GPT-5.6-Cyber, a specialized model achieving 95% completion on advanced exploit tasks. Daybreak platform expanded into Blue (defender) and Red (offensive) tiers. The model found real-world vulnerabilities in Chrome V8. This marks a significant step in AI-driven cybersecurity with safety controls.