en.Wedoany.com Reported - On May 12, 2026, OpenAI officially launched Daybreak, a software security defense project for enterprises, bringing cyber defense capabilities forward into the software development phase. It uses AI to automatically discover high-risk vulnerabilities and provide assessment and remediation support. The project integrates the GPT-5.5 model series, the Codex agent execution framework, and an ecosystem of 26 security vendors, aiming to build defense capabilities from the code-writing stage rather than patching vulnerabilities after deployment.
Daybreak offers a three-tier model architecture, corresponding to different risk levels and authorization scenarios. The standard GPT-5.5 is for general-purpose use, equipped with conventional security protection mechanisms; GPT-5.5 with Trusted Access for Cyber is specifically for authenticated defense teams, covering workflows such as secure code review, vulnerability triage, malware analysis, detection engineering, and patch verification; GPT-5.5-Cyber is available as a limited preview, suitable for authorized red team exercises, penetration testing, and controlled validation, with stricter account-level controls. OpenAI CEO Sam Altman stated that AI is already very powerful in the cybersecurity field and is about to become even stronger, and the company hopes to collaborate with as many enterprises as possible to continuously help them defend against cyber threats.
In terms of workflow, Daybreak reads enterprise software code repositories through Codex Security, automatically generates editable threat models, and then continuously monitors high-risk vulnerabilities. Once an issue is discovered, enterprises can conduct in-depth investigations in an isolated environment, forming a continuous closed loop of "modeling, monitoring, and investigation." The project is not a one-time vulnerability scanning service but a security toolchain embedded in daily development. Development teams can complete security checks within the daily cycle of writing code, modifying code, and verifying patches, compressing vulnerability analysis that originally took hours into minutes.
The project is built upon GPT-5.4-Cyber, which OpenAI released in April 2026 and which has already helped fix over 3,000 vulnerabilities. Codex Security, the core execution framework of Daybreak, was first released in March 2026, featuring capabilities in code repository scanning, threat modeling, vulnerability verification, and automated remediation.
On the ecosystem partnership level, OpenAI has established collaborations with over 26 security vendors, including Cloudflare, Cisco, CrowdStrike, Palo Alto Networks, Oracle, Akamai, and Intel, covering the complete chain from vulnerability discovery, patch remediation, and threat monitoring to software supply chain security. Currently, enterprises can apply to OpenAI for a Daybreak assessment, which includes a comprehensive vulnerability scan, but pricing has not yet been announced. OpenAI also stated that it is working with industry and government partners to prepare for the deployment of even stronger cybersecurity models in the future.
Previously, Anthropic launched Project Glasswing in April 2026, offering cyber defense services to enterprises based on its unreleased Claude Mythos model. Mozilla disclosed that it used Mythos to discover and fix 271 vulnerabilities in the latest version of Firefox. Apple, Microsoft, Google, and Amazon have already adopted Glasswing, intensifying the competition between the two AI companies in the enterprise security market. OpenAI's Daybreak launch, with its three-tier authorization architecture and embedded development toolchain as a differentiating approach, directly responds to Anthropic's positioning in the same arena.
This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com
