OpenAI's Daybreak suite uses GPT-5.5 for automated vulnerability detection

OpenAI has unveiled Daybreak, a cyber defense suite that combines its latest GPT-5.5 large language model with the Codex coding tool to automatically identify and remediate software vulnerabilities. The suite offers three model tiers: a default GPT-5.5 for general-purpose security tasks, GPT-5.5 with Trusted Access For Cyber for defensive workflows like secure code review and malware analysis, and GPT-5.5-Cyber for authorized red teaming and penetration testing. Daybreak’s capabilities include threat modeling, patch validation, and dependency risk analysis.

OpenAI is not making Daybreak publicly available, instead partnering with industry leaders like Cloudflare, Cisco, Oracle, and Akamai. Pricing remains undisclosed, with interested parties directed to OpenAI’s sales team. This controlled rollout mirrors Anthropic’s recent Project Glasswing and Mythos initiative, which similarly restricts access to powerful AI-driven security tools. The move underscores a growing trend among AI labs to limit distribution of models that could pose dual-use risks in cybersecurity.