OpenAI is launching Daybreak, an AI initiative focused on detecting and fixing vulnerabilities before attackers find them. Daybreak uses the Codex Security AI Agent, launched in March, to build a threat model based on an organization’s code and focus on potential attack paths, validate potential vulnerabilities, and then detect high-risk ones.
Its launch comes just a month after rival Anthropic announced Cloud Mythos, a security-focused AI model that it claimed was too dangerous to release publicly and was only shared privately as part of its own initiative, called Project Glasswing. Nevertheless, this did not prevent at least some unauthorized parties from gaining access.
However, OpenAI lacks a similar security product until now. Like Glasswing, Daybreak isn’t built on just one AI model – OpenAI says that “Daybreak brings together the most capable OpenAI models, codecs, and our security partners.”
Daybreak also includes special cyber models, including GPT-5.5 with Trusted Access for Cyber and GPT-5.5-Cyber, which debuted last week. OpenAI also says it is working with its “industry and government partners” as it prepares to “deploy more cyber-enabled models.”
<a href