openai has launched the daybreak project, positioning itself to compete with anthropic’s glasswing initiative and the mythos ai model. according to foreign media reports, openai officially unveiled daybreak today, aiming to rival anthropic’s glasswing program and the mythos ai model. like glasswing, daybreak is a specialized cybersecurity project designed to help tech companies identify and address security vulnerabilities in their platforms.
the core objective of daybreak is to embed cybersecurity defenses directly into the software development process. built upon openai’s gpt-5.4-cyber security large model—upgraded and released in april this year—the company claims this model has already helped patch over 3,000 security vulnerabilities.
openai ceo sam altman stated that the company hopes to collaborate with as many enterprises as possible, helping them continuously strengthen their software systems and defend against various cyber threats. currently, companies such as apple, microsoft, google, and amazon have already joined anthropic’s glasswing initiative.
powered by codex’s security capabilities, daybreak can generate editable threat models based on an organization’s software code repositories and automatically conduct routine monitoring for high-risk vulnerabilities. any identified security risks can then be thoroughly traced and analyzed within an isolated environment.
enterprises can apply to openai for daybreak’s security assessment services, which include vulnerability scanning features. pricing information has not yet been disclosed by the company.
