OpenAI Introduces GPT-6 Cyber for Advanced Security Research

OpenAI has released GPT-6 Cyber, a cybersecurity-focused model designed to help security researchers identify vulnerabilities, assess how they could be exploited and develop fixes before malicious actors take advantage of them.

The model is initially being made available to a small group of vetted customers. OpenAI classified it as “Critical” for cybersecurity under its safety framework, reflecting the advanced capabilities associated with vulnerability discovery and exploit development.

The launch comes as the company reports increasing attempts to misuse artificial intelligence, including activity by malicious cyber actors and groups seeking to distill its models.

OpenAI’s approach is to make the model available for authorized security work while limiting access to approved users.

Vulnerability Research Demonstrates Model Capabilities

OpenAI reported that its researchers used GPT-6 Cyber to identify 81 previously unknown vulnerabilities in Firefox. Mozilla classified the findings as high severity and has begun releasing fixes.

The model also helped develop working exploits for 948 Linux kernel vulnerabilities, including 588 that OpenAI said were previously unknown.

In addition, the company disclosed 304 critical Linux findings to maintainers, together with proposed patches.

These results illustrate the model’s intended role in security research: helping specialists identify weaknesses and evaluate potential attack methods so that vulnerabilities can be addressed before they are exploited in real-world attacks.

Cybersecurity Classification Raised to Critical

GPT-6 Cyber carries a higher cybersecurity classification than its predecessor. OpenAI assigned the new model a “Critical” rating, compared with the “High” rating given to GPT-5.6 Cyber.

The distinction reflects the company’s assessment of the model’s cybersecurity capabilities and the need to manage access carefully.

OpenAI has also highlighted growing attempts to misuse AI systems. The company cited malicious cyber activity and efforts by groups to distill its models as part of the broader security environment surrounding advanced AI development.

Access Limited to Approved Daybreak Red Customers

GPT-6 Cyber is being distributed through Daybreak Red, OpenAI’s access tier for advanced and authorized cybersecurity activities, including penetration testing and exploit development.

The initial release is restricted to select customers. OpenAI said access would expand to all approved Daybreak Red customers over the coming weeks.

At the same time, the company is making its earlier GPT-5.6 Cyber model available to all Daybreak Red customers starting Thursday.

This staged approach allows OpenAI to extend access to cybersecurity professionals while maintaining approval requirements for the use of its advanced security capabilities.

What Investors and the Cybersecurity Sector Should Watch Next

The release of GPT-6 Cyber highlights the growing role of advanced AI in vulnerability discovery and defensive security research. Its reported findings across Firefox and the Linux kernel demonstrate the potential for AI systems to accelerate the identification and remediation of software weaknesses.

The next developments to watch include the expansion of access to approved customers, the practical use of the model in authorized security environments and how OpenAI manages the risks associated with increasingly capable exploit-development tools.

The balance between defensive applications and potential misuse will remain central to the adoption of advanced AI in cybersecurity.