OpenAI has launched Daybreak, a new cybersecurity initiative developed as a direct competitor to Anthropic’s Project Glasswing initiative. This new move aims to make cyber defense processes faster and more efficient with artificial intelligence models.

Anthropic’s Claude Mythos Preview model recently helped Mozilla detect and patch 271 vulnerabilities in the Firefox browser. OpenAI, on the other hand, aims to achieve a similar success rate by using its own private security agent Codex and its current models within the scope of Daybreak.

A New Approach to Cyber ​​Defense

OpenAI bases Daybreak’s core philosophy on integrating cyber defense from the very beginning of the software development process. The company states that this system will not only find and fix existing vulnerabilities, but also transform processes into a proactive structure.

Daybreak reduces hours of analysis into minutes by prioritizing high-impact security issues. The system can automatically create and test patches within code repositories and send the results back to customers’ systems with auditable evidence.

In an example scenario, Codex Security scans a codebase, verifying the highest-risk findings, and then automatically fixes those issues. This method aims to significantly ease the workload on security teams.

GPT-5.5 Supported Security Workflows

The Daybreak initiative is powered by GPT-5.5, one of OpenAI’s newest models. While standard GPT-5.5 is used for general-purpose operations, GPT-5.5 with Trusted Access for Cyber ​​comes into play in defense-oriented security workflows.

This particular model; It is used in critical tasks such as secure code review, vulnerability triage, malware analysis and patch verification. Additionally, the GPT-5.5-Cyber ​​model is available with preview access for specialized workflows such as authorized penetration testing and controlled verification.

OpenAI is collaborating with leading names in the industry for this new initiative. Giant companies such as Cloudflare, Cisco, CloudStrike, Palo Alto Networks, Oracle and Akamai are among the first partners to start working with OpenAI as part of Daybreak.

This wide partnership network is considered an important step for the adoption of the initiative at the corporate level.

How effective do you think such artificial intelligence-supported cyber security initiatives can be in closing vulnerabilities in the software world?