OpenAI Launches Daybreak to Automate Cyber Defense and Software Patching

OpenAIOpenAI

· Updated

OpenAI launched Daybreak, a cybersecurity platform that integrates GPT-5.5 models and Codex to identify, validate, and remediate software vulnerabilities. The system shifts security teams from manual triage to autonomous agentic workflows that can generate and test code patches directly in repositories.

OpenAI launched Daybreak, a cybersecurity platform that unifies its frontier models with the Codex agentic coding harness. The system automates the security flywheel by reasoning across codebases to find vulnerabilities, generating fixes, and verifying remediation. It includes GPT-5.5-Cyber, a version with permissive safeguards for authorized penetration testing.
Core models
GPT-5.5 and GPT-5.5-Cyber
Agentic harness
Codex Security
Access tiers
Standard, Trusted Access, and Cyber
Key capabilities
Triage, patch generation, and verification
Availability
Vulnerability scan request or Sales contact

This launch shifts AI from a reactive assistant to an active defender. By leveraging the agentic capabilities of Codex Security, Daybreak aims to reduce the time between threat discovery and remediation from hours to minutes. This vertical expansion mirrors OpenAI's GPT-Rosalind reasoning model, arming defenders against AI-accelerated attackers.

You can now request a vulnerability scan or contact sales to deploy the platform. Access is tiered: standard GPT-5.5 for general work, GPT-5.5 with Trusted Access for Cyber for defensive workflows, and the restricted GPT-5.5-Cyber for specialized testing. OpenAI plans to roll out more capable models to partners soon.

View the full update on openai.com
OpenAI
OpenAI
@OpenAI
X

Introducing Daybreak: frontier AI for cyber defenders. Daybreak brings together the most capable OpenAI models, Codex, and our security partners to accelerate cyber defense and continuously secure software. A step toward a future where security teams can move at the speed defense demands.

1.2kretweets11klikes
View on X

Still wondering? A few quick answers below.

OpenAI Daybreak is a cybersecurity platform designed to help defenders secure software and systems. It combines OpenAI's frontier models, such as GPT-5.5, with the Codex agentic harness to identify vulnerabilities, generate code patches, and verify remediation. The goal is to move security teams from manual analysis to automated, resilient-by-design workflows.

Daybreak uses AI to reason across codebases and analyze unfamiliar systems to find subtle risks. It leverages Codex as an agentic harness—a system that allows the AI to take multi-step actions—to generate and test patches directly in repositories. It also provides audit-ready evidence to verify that security issues have been properly fixed.

Daybreak is intended for cybersecurity defenders, including developers, security teams, and government partners. While standard GPT-5.5 is available for general use, specialized versions like GPT-5.5-Cyber are restricted to verified institutions and authorized workflows. Interested organizations can request a vulnerability scan or contact the OpenAI sales team to align on the best model for their needs.

OpenAI offers three access levels for Daybreak. Standard GPT-5.5 has general-purpose safeguards. GPT-5.5 with Trusted Access for Cyber provides more precise safeguards for verified defensive work like malware analysis. GPT-5.5-Cyber is the most permissive version, designed for specialized authorized workflows such as red teaming and penetration testing, paired with strict account-level controls.

Organizations can request a vulnerability scan directly through the OpenAI Daybreak website. This service allows teams to see where OpenAI's models can help prioritize, validate, and fix security issues. By using these scans, defenders can reduce hours of manual analysis to minutes, focusing their efforts on the high-impact threats that matter most to their infrastructure.

Every HeadsUpAI update is written based on its original source and reviewed before it's published. Read our editorial standards →

OpenAI

See all AI news & updates from OpenAI →

Share this update