OpenAI Launches Two-Tier Security Access Program Alongside GPT 5.6 Cyber

OpenAI announced GPT-5.6-Cyber and a two-tier Daybreak program (Blue for defensive work, Red for advanced defensive and offensive tasks), describing improved cybersecurity task performance and reduced refusals; the company says the model helped find and responsibly disclose a high-severity Chrome V8 vulnerability (CVE-2026-15903) while highlighting guardrail and deployment concerns.