OpenAI launches Daybreak in response to Anthropic's Glasswing cybersecurity initiative based on Claude Mythos

OpenAI launches the Daybreak project in cybersecurity

OpenAI announced the launch of a new project called Daybreak, which aims to integrate protection tools directly into software rather than simply searching for and fixing vulnerabilities after they are discovered. This is a clear response to Anthropic’s similar initiative called Glasswing.

What does Glasswing offer?
* Anthropic customers gain access to the advanced AI model Claude Mythos Preview.
* Success example: Mozilla used this model to identify and close 271 vulnerabilities in the latest version of the Firefox browser.

Key features of Daybreak
The architecture implements a variety of AI models, including several versions of GPT‑5.5 and a special Codex agent. Trusted Access for Cyber allows GPT‑5.5 to be used with elevated access rights for security tasks. GPT‑5.5‑Cyber provides “preview access” to specialized workflows: automated threat analysis, penetration testing, and controlled checks. Reducing analysis time – the goal is to cut the time from problem detection to resolution from hours to minutes. Patch generation and testing – automatically creates and verifies updates in repositories, then sends results back to client systems.

Project partners
OpenAI already collaborates with the following companies:
- Cloudflare
- Cisco
- CloudStrike
- Palo Alto Networks
- Oracle
- Akamai

These partnerships help integrate Daybreak into existing infrastructures and expand its capabilities.

Conclusion
Daybreak seeks to turn cyber defense from a reactive process into a proactive tool embedded directly in software. This will allow companies to respond faster to threats, reducing the time from detection to vulnerability remediation and improving overall system security.