Newshot OpenAI launches two AI cybersecurity tools to automate penetration testing and vulnerability remediation
OpenAI has officially released two AI cybersecurity tools, Daybreak Red Team and Daybreak Defense Factory, to help organizations of all kinds test web applications, confirm security flaws, and develop remediation plans.
Announced on November 12, 2026, the two products are feature extensions of the OpenAI Daybreak platform that connect automated penetration testing with vulnerability remediation workflows, while approval authority over all code changes remains with engineers.
The release is part of OpenAI's "Cyber at the Frontier" event. Together, the two tools target a widespread security pain point in the industry: discovering a potential vulnerability is only the starting point of security work. Teams also need to determine whether an attacker could exploit the flaw, assess its real-world impact, and verify whether a proposed patch actually eliminates the risk.
Red Team opens early access, can verify real vulnerability risk
Daybreak Red Team, currently in early access, uses OpenAI models to detect vulnerabilities in web applications. According to Nancy Wang, head of AI for cyber applications at OpenAI, the service combines continuous scanning capabilities with agents to examine source code, application context, and potential attack paths.
The tool does not stop at outputting scan alerts; its core goal is to verify whether reported issues constitute a real security risk. It can also process findings submitted by code scanners, bug bounty programs, and vulnerability disclosure reports, and test whether patches designed for those vulnerabilities are effective.
The security workflow OpenAI designed for it connects vulnerability discovery with actual verification, confirming whether a vulnerability is reachable and exploitable in a real application before remediation priorities are set.
Defense Factory generates remediation patches with proof of effectiveness testing
Daybreak Defense Factory focuses on vulnerability discovery, verification, and remediation across the entire codebase workflow. Its agents use OpenAI cybersecurity-specific models, including GPT-6 Cyber, to investigate potential vulnerabilities and generate proposed patches; all patches must be reviewed and approved by engineers before going to production.
Engineers receive a merge-ready pull request containing the patch code along with test results proving the fix is effective. Compared with isolated security alerts, this output format is more practical: developers get clear remediation recommendations and can evaluate the supporting verification evidence during code review.
Red Team identifies application weaknesses and attack paths, while Defense Factory turns confirmed vulnerabilities into proposed code fixes, helping security researchers and developers address vulnerabilities faster.
The feature update continues OpenAI's previous technical direction. As Cybersecurity News previously reported, the agent in OpenAI Codex Security can build project-specific security context and verify the authenticity of findings in an isolated environment. Once verification passes, the agent generates a corresponding patch for reviewers to examine.
This update also extends the capabilities of the Codex Security Cloud service. Codex Security Cloud is an always-on security service that monitors code repositories and all incoming commits, automatically investigating potential vulnerabilities. It deduplicates findings and then generates remediation plans, reflecting OpenAI's push to embed security capabilities continuously into development workflows.
Retaining human approval boundaries, embedding into existing development security workflows
The new features clearly define the boundaries of human approval: Defense Factory only generates proposed changes and does not deploy them independently; Red Team is currently in early access with limited availability.
The practical value of such tools depends on whether they can provide teams with reliable verification evidence and practical patch proposals while reducing invalid alerts that require lengthy manual investigation.
OpenAI positions these two products as completing the full security workflow rather than simply increasing the number of reported vulnerabilities. According to its cybersecurity platform description, this end-to-end process from discovery to verification is designed for authorized defenders and can adapt to enterprises' existing security and development workflows.
OpenAI Launches Two AI Cybersecurity Tools to Automate Penetration Testing and Vulnerability Remediation