root@ransomvault:~$ tail -f threat_intel.log

OpenAI Launches Daybreak Red Team and Defense Factory: AI Agents for Pentesting and Vulnerability Patching

Darkintel Team2 min read

OpenAI has announced two new AI-powered security products: Daybreak Red Team and Daybreak Defense Factory. They were unveiled at the company’s Cyber at the Frontier event on November 12, 2026. Red Team focuses on penetration testing and vulnerability validation, while Defense Factory investigates security findings and prepares patches.

Daybreak Red Team: AI-driven penetration testing
Daybreak Red Team uses multiple AI agents to find and investigate potential security weaknesses in web applications. According to OpenAI, the agents use source code and application context to investigate attack paths and confirm vulnerabilities.

Beyond its own testing, the product can examine findings from:

  • Code scanners
  • Bug bounty submissions
  • Vulnerability disclosure reports

To keep testing safe, OpenAI says the agents run in isolated environments. A feature called Guardian review enforces testing boundaries and controls which applications and endpoints the agents can access.

Daybreak Defense Factory: from finding to fix
Defense Factory is built to help organizations fix vulnerabilities found by their security tools. OpenAI says its agents trace vulnerabilities through an organization’s code and develop patches, which engineers then review and test before merging.

Key capabilities include:

  • Investigating findings from Daybreak Red Team and other security products
  • Importing vulnerability data from Wiz, Snyk, and Endor Labs
  • Scanning connected GitHub repositories, monitoring new commits, and running scheduled security assessments
  • Letting engineers review supporting evidence, edit proposed patches with Codex, and open draft GitHub pull requests

Importantly, patches are not deployed automatically. Human engineers stay in control of what gets merged.

At a glance

Daybreak Red TeamDaybreak Defense Factory
PurposePenetration testing and vulnerability validationVulnerability investigation and patch development
ScopeWeb applicationsConnected GitHub repositories
InputsSource code, app context, scanner results, bug bounty and disclosure reportsFindings from Daybreak Red Team, Wiz, Snyk, Endor Labs, and other tools
OutputConfirmed vulnerabilities and attack pathsProposed patches as draft GitHub pull requests
SafeguardsIsolated environments and Guardian reviewEngineer review and testing before merge; no automatic deployment
AvailabilityEarly access by requestPreview plugin in ChatGPT for workspaces with Daybreak access

Why it matters
Daybreak marks OpenAI’s move into offensive and defensive security tooling. It joins a growing trend of AI agents being used to find and fix vulnerabilities at scale, a trend we have seen repeatedly in recent disclosures, where AI-assisted research played a role in finding critical flaws in widely used software. Tools like these could help security teams shrink large vulnerability backlogs and shorten the time from discovery to fix. They also raise important questions about testing boundaries, the accuracy of AI-generated patches, and how much access autonomous agents should have to production code and applications.

Availability
OpenAI is accepting early access requests for Daybreak Red Team. Defense Factory is available as a preview plugin in ChatGPT for workspaces with Daybreak access.

Considerations for security teams

  • Keep humans in the loop: Treat AI-generated patches like any other code contribution, with full code review, testing, and approval before merging.
  • Define testing scope carefully: Make sure only authorized applications and endpoints are in scope for automated penetration testing.
  • Validate findings: Cross-check AI-confirmed vulnerabilities against your existing tools and manual review to manage false positives and false negatives.
  • Review data access: Understand what source code, repository, and vulnerability data the agents can reach, and make sure it fits your data governance policies.

Learn more: Daybreak Red Team | Daybreak Defense Factory

Newsletter

Get the latest cyber news in your inbox

Ransomware, breaches and exploited vulnerabilities, delivered by RansomVault.

Free. No spam. Unsubscribe in one click.

Latest news

How PHP2Shell Works, and Why It Gives Attackers Access Before Your Code Even Runs

Knowledge

PHP2Shell, the pair of critical PHP vulnerabilities tracked as CVE-2026-109618 and CVE-2026-109619, has become one of the most heavily exploited bugs of the year. It was exploited four days before PHP released patches on October 21, added to CISA’s Known Exploited Vulnerabilities catalog, and most recently used to break into Check Point Security Gateways and wipe VMware ESXi hosts. But how does it actually work? This post explains the vulnerability in plain terms, using technical research on th

FAA Grounds All U.S. Flights Over Nationwide Air Traffic Control Communications Failure

Threat Intelligence

On Friday afternoon, the Federal Aviation Administration (FAA) ordered a nationwide ground stop at every U.S. airport, citing "extraordinary circumstances" and safety concerns. U.S. Transportation Secretary Sean Duffy said a communications problem was affecting all Air Route Traffic Control Centers across the country. "FAA technicians are investigating the issue. We will provide immediate updates as we have them," Duffy wrote on X. Departing flights must stay on the ground until further noti

PHP2Shell Attacks on Check Point Gateways Deploy Data Wiper on VMware ESXi Hosts

Zero-Day / Threat Intelligence

Threat actors are exploiting vulnerable Check Point Security Gateways in a campaign that ends with data-wiping malware on VMware ESXi hosts. According to Check Point Research, the attacks began on October 30, 2026, first against Israeli organizations. A single actor ran the whole operation, from initial access to lateral movement and destruction. Background: PHP2Shell PHP2Shell is the name for two critical PHP vulnerabilities, CVE-2026-109618 and CVE-2026-109619, that allow unauthenticated remo