Rhymes with Mythos.
By Stan Schroeder on May 12, 2026.
OpenAI has introduced Daybreak, a suite of cyber defense tools leveraging its LLMs to detect and fix software vulnerabilities. Currently, the tools are not widely available; instead, OpenAI plans collaboration with industry and government partners to gradually implement more cyber-capable models.
Daybreak harnesses the power of OpenAI models, particularly the new GPT-5.5, combined with the flexibility of Codex, OpenAI’s coding tool, for tasks like secure code review, threat modeling, and remediation guidance.
Three models are available: GPT-5.5 for general work, GPT-5.5 with Trusted Access For Cyber for defensive operations, and GPT-5.5-Cyber for red teaming, penetration tests, and controlled validation.
Inspired by initiatives like Anthropic’s Project Glasswing and Mythos, Daybreak restricts access to selected partners, with OpenAI withholding public pricing. Companies like Cloudflare, Cisco, Oracle, and Akamai are among the named partners. Interested parties are encouraged to contact OpenAI’s sales team for pricing details.