OpenAI just launched Daybreak, its most ambitious cybersecurity initiative to date. Built on GPT-5.5 and the Codex Security agent, this silver form it promises to detect and fix flaws in the software before an attacker has the opportunity to exploit them. And comes directly into competition with Anthropic.

What is OpenAI Daybreak and how it works

The key here is that Daybreak is not just an AI model, but a complete cyber defense system. It combines the intelligence of the G models PT-5.5 with the capabilities of the Codex Security agent, which builds an editable threat map from a company's code repository. 

That map focuses on realistic attack paths and higher risk code, identifies vulnerabilities in an isolated environment and then proposes solutions What this means is that the system acts like an autonomous security team that works continuously, not only when something has already gone wrong. 

The capabilities include, among others:

The three models of GPT-5.5 Cyber offered Daybreak

OpenAI didn't launch a single model but three levels of access, depending on the type of work and the user's verification level. This is what differentiates them: 

According to OpenAI, its previous GPT-5.4-Cyber ​​model has already contributed to fixing more than 3,000 vulnerabilities. Daybreak is, in a nutshell, the natural evolution of that effort. 

OpenAI vs Anthropic in the race for cybersecurity with AI

We are in direct competition between the two most influential AI companies in the sector. Anthropic recently launched Project Glasswing and the Claude Mythos model, also focused on cybersecurity. OpenAI responded weeks later with Daybreak. 

The most noticeable difference is that, while Glasswing operates only with Claude Mythos, Daybreak offers a family of three models with different levels access and verification elements. That gives more flexibility to companies to choose what level of capacity they need based on their type of operation. 

Sam Altman stated that OpenAI wants to work with“as many companies as possible” to achieve continuous software security. The bet is integrating cyber defense from the beginning of the development cycle, not as a final patch. 

What is OpenAI Daybreak?OpenAI Daybreak is a cybersecurity platform with artificial intelligence launched in May 2026. Combines GPT-5 models .5 and GPT-5.5-Cyber with Codex Security agent to detect, analyze and repair vulnerabilities in software before they are exploited by attackers. 

What is the difference between GPT-5.5 and GPT-5.5-Cyber? GPT-5.5 is the standard model for general use, while GPT-5.5-Cyber is the more advanced and restricted version designed for specialized eas such as network teaming and penetration testing in authorized environments. Access to GPT-5.5-Cyber requires account verification and is available in limited preview. 

How is OpenAI Daybreak different from Anthropic's Project Glasswing? Both are AI-based cybersecurity initiatives, but Daybreak offers three levels of model with different access controls , while Glasswing operates exclusively with Claude Mythos. Daybreak also integrates agent capabilities through Codex Security to build threat models from the code repository. 

The race to dominate cybersecurity with AI is just beginning. If you're interested in knowing how these tools can protect the software you use daily, keep exploring our analysis on the latest news from OpenAI and Anthropic.