OpenAI launches GPT-5.6-Cyber for cybersecurity, expands Daybreak into two tiers
OpenAI split Daybreak into Blue and Red tiers and released GPT-5.6-Cyber built on GPT-5.6 Sol, restricted to Red-tier trusted customers.
On August 11 local time, OpenAI announced an expansion of its Daybreak cybersecurity program, splitting participation into Blue and Red tiers and releasing a new model GPT-5.6-Cyber built specifically for cybersecurity tasks.
Daybreak is a cybersecurity partnership program OpenAI first launched in May this year, aimed at helping ecosystem partners use the company's most advanced models to address evolving cyber threats. Under the expansion, Blue tier is for organizations conducting defensive security work, with access to OpenAI's advanced general models with safety guardrails partially adjusted for defensive scenarios to identify risks, analyze threats, and complete related security tasks.
Red tier offers higher-level access for participants to use specialized cybersecurity models for security testing, vulnerability research, and exploit verification. Released alongside Red, GPT-5.6-Cyber is built on OpenAI's most powerful publicly available GPT-5.6 Sol model and is optimized for professional cybersecurity tasks. The model achieves 95% completion on advanced cybersecurity benchmarks, far above the 1.5% of general models, and has helped discover high-severity vulnerabilities including Chrome V8.
GPT-5.6-Cyber is currently available only to trusted customer partners, with initial access granted to Accenture, IBM, CrowdStrike, and Cloudflare. OpenAI said in its announcement it hopes to put frontier intelligence in the hands of trusted defenders before attackers deploy offensive AI at scale. The expansion comes amid heightened attention to AI safety incidents, with OpenAI, Anthropic, and Meta disclosing cybersecurity testing incidents in recent weeks.