GPT-6 Cyber: 81 Firefox Vulnerabilities — But Access Is Limited to Specialists

TVS News AI, editorially supervised, presents GPT-6 Cyber: 81 Firefox Vulnerabilities — But Access Is Limited to Specialists.

81 previously unknown security vulnerabilities in Firefox — that is the concrete number OpenAI’s new security AI has already provided as a readily measurable success. GPT-6 Cyber is not simply a general-purpose chatbot for IT questions; it is designed to search for security vulnerabilities and suitable fixes.

The key limitation, however, lies in access: for now, GPT-6 Cyber remains a kind of protected specialist tool. That means you cannot simply install the technology yourself or use it for your own projects. Access remains reserved for cybersecurity specialists.

As a result, the announcement is aimed less at end users than at security teams, developers, and companies that want their systems tested. OpenAI’s use of the model with Firefox also demonstrates what it is designed to do: find as many vulnerabilities as possible, suggest appropriate fixes, and secure them before broader use.

GPT-6 Cyber Discovers 81 Serious Firefox Flaws

81 severe vulnerabilities have been found in Firefox using the model, according to the report. The number of findings is not the only remarkable aspect. OpenAI explicitly states that GPT-6 Cyber identifies previously unknown security vulnerabilities and therefore does not search exclusively for issues that are already known or publicly documented.

For cybersecurity, that is a significant difference. A known vulnerability can be checked against a list. An unknown vulnerability, by contrast, must first be found, understood, and assessed. That is precisely where OpenAI’s solution comes in — at the transition from discovering a vulnerability to preparing a technical fix.

The report does not name the individual vulnerabilities or compare the model with conventional software. It does, however, show a concrete real-world application: GPT-6 Cyber has already been tested against Firefox and, according to the report, identified 81 serious vulnerabilities.

Finding Vulnerabilities and Suggesting Fixes Are Two Different Tasks

Suggesting fixes — that is the model’s second central capability. It goes beyond simple error hunting. The model is not supposed merely to report where a problem lies, but to offer a concrete technical path to fixing it.

That does not, however, mean every vulnerability found would be automatically fixed. A proposed fix must be reviewed, tested, and integrated into the relevant environment. The sources report only on the ability to suggest fixes; they do not claim that GPT-6 Cyber independently deploys changes in Firefox without any form of oversight.

This is exactly where the practical benefit lies if you want to assess security vulnerabilities quickly. The combination of detection and actionable options makes the model more interesting to specialists than pure analysis that ultimately produces nothing more than a list of errors.

Why GPT-6 Cyber Remains Under Lock and Key for Now

The most striking limitation is not the model’s performance, but access: GPT-6 Cyber is currently reserved exclusively for cybersecurity specialists. The tool therefore remains limited to a clearly defined group of users rather than being available in app stores or obtainable as an ordinary software download.

For the public, there is a clear practical consequence: according to the current report, it is not possible to test it with Firefox, use it on your own systems, or download it for private applications. Anyone who wants to use GPT-6 Cyber must turn to the authorized specialists.

This limitation does not diminish the technology that has been introduced. It merely shows that the launch is being deliberately restricted. The report names neither a price nor a release date for general access. Nor does it mention a public testing program. There is no basis in the available information for drawing conclusions about when or under what conditions the restriction will be lifted.

What the Firefox Deployment Reveals About the Security AI

The deployment with Firefox turns GPT-6 Cyber into more than a theoretical model announcement. OpenAI presents 81 severe vulnerabilities found as a tangible result. At the same time, the restricted release shows that security AI is being used with sensitive responsibilities: vulnerabilities reported by such a system can have serious consequences.

The available information, however, contains only the points already mentioned: detecting previously unknown vulnerabilities, suggesting fixes, use with Firefox resulting in 81 serious findings, and currently specialized access. The report provides no detailed information about the individual Firefox findings, the model’s technical operation, or a wider release.

Conclusion: Impressive Capabilities, Deliberately Narrow Access

GPT-6 Cyber demonstrates how far OpenAI has already advanced automated vulnerability hunting. The combination of previously unknown vulnerabilities, proposed fixes, and 81 hits in Firefox makes the announcement concrete — and unusually relevant to cybersecurity.

At the same time, open access is not on the agenda: for now, GPT-6 Cyber remains reserved for specialist professionals. Anyone who wants to benefit from its capabilities therefore depends on specialized access. In future, the model’s greatest value to the broader IT world may not lie in direct use, but in the vulnerabilities it uncovers before they are made public.