OpenAI Launches ‘Daybreak’ to Help Build Secure By Design Software

https://www.infosecurity-magazine.com/news/openai-daybreak-secure-by-design/

Publish Date: 2026-05-28 02:49:39

Source Domain: www.infosecurity-magazine.com

OpenAI introduces Daybreak for secure software development

OpenAI has launched Daybreak, its new initiative leveraging frontier large language models (LLMs) and its AI-coding assistant Codex to assist developers in constructing secure software throughout the entire development lifecycle. Unveiled on May 12, Daybreak builds upon the Trusted Access for Cyber security program, which limits access to certain advanced models to vetted organizations. Early components include three advanced models, notably GPT-5.5, and an exclusive code-review assistant called Codex Security. The initiative aims to compress manual security analysis by helping defenders build threat models, discover and test vulnerabilities swiftly in isolated environments, and validate patches. However, cybersecurity experts warn that organizations must carefully govern access provided to these AI tools to prevent exposure to sensitive data, highlighting the need for robust patch management systems as the speed of vulnerability discovery increases.

Key Points:

OpenAI’s Daybreak aims to use advanced LLMs to secure software from inception, leveraging models like GPT-5.5 and Codex.
The initiative involves three advanced models, including a specialized version offering security safeguards, and Codex Security for automated vulnerability detection.
While Daybreak promises efficiency in security analysis, experts caution about the data governance risks and challenges in patch management as AI-assisted discovery accelerates.
The Trusted Access for Cyber (TAC) program includes hundreds of organizations, emphasizing cybersecurity and large enterprises, with plans for governmental expansion.