OpenAI has made its specialized GPT-6 Cyber model available to selected customers in the Red tier of its Daybreak cybersecurity program. The Red tier, launched on August 10 and 11, 2026, is restricted to authorized vulnerability research and security testing, while the Blue tier covers standard defensive cybersecurity work. According to the article, GPT-5.6 Cyber successfully handles 95% of complex cybersecurity prompts, compared with 1.5% to 2% for the standard model. The model has already identified previously unknown vulnerabilities in real-world software, including two critical flaws in Chrome’s V8 engine, among them CVE-2026-15903. Access to Daybreak requires identity verification and legal attestations, and OpenAI maintained these strict controls when the program expanded to AWS GovCloud in October 2026.
Source: Read the original article
