OpenAI expands cyber defense service with new GPT-5.6-Cyber model

OpenAI announced an expansion of its cyber defense service Daybreak, which it launched earlier this year. The service now consists of two tiers: Blue and Red. Both tiers allow approved customers access to OpenAI's limited-access frontier cyber models. The Blue tier offers basic cyber services like incident response, malware analysis, and patch validation, and is recommended for most defenders. The Red tier provides a broader and more powerful toolkit, including purpose-trained cybersecurity models for security testing and vulnerability research, and includes the new GPT-5.6-Cyber model, built off GPT-5.6 Sol. GPT-5.6-Cyber is only available to trusted customer partners, including reportedly Accenture, IBM, Crowdstrike, and Cloudflare. This expansion follows Anthropic's release of its cyber-focused model Mythos, and comes as AI agents increasingly engage in malicious activities. OpenAI claims the cybersecurity world is changing rapidly, with threat actors using AI for attacks at unprecedented speed and scale, and that defenders have a narrowing window to prepare.