OpenAI GPT-6 Cyber: Features and Access

Published November 6, 2026

OpenAI has launched GPT-6 Cyber, a specialized artificial intelligence model designed to discover software vulnerabilities, test their potential impact and help security teams develop fixes.

Announced on November 5, 2026, the model represents a major advance in AI-assisted cybersecurity research. According to OpenAI, it has already helped uncover 81 previously unknown high-severity vulnerabilities in Firefox and develop working exploits for 948 Linux kernel vulnerabilities.

Unlike a typical ChatGPT model, GPT-6 Cyber is not available to the general public. Access is initially limited to approved cybersecurity customers.

What Is GPT-6 Cyber?

GPT-6 Cyber is OpenAI’s latest model designed for advanced security research.

It can assist authorized researchers with several tasks:

Discovering previously unknown software vulnerabilities.
Validating whether security weaknesses can be exploited.
Developing and testing exploits in authorized environments.
Analyzing complex software codebases.
Helping researchers create and verify security patches.

The model is designed to perform these tasks while respecting the boundaries established by security researchers.

This is particularly important because advanced cybersecurity AI can potentially be used for both defensive research and malicious attacks.

GPT-6 Cyber Found 81 Firefox Vulnerabilities

One of the most significant findings announced by OpenAI involves Mozilla Firefox.

OpenAI says its researchers used GPT-6 Cyber to identify 81 previously unknown vulnerabilities that Mozilla classified as high severity.

The model also provided reproducible evidence and helped propose potential fixes.

According to OpenAI, Mozilla has already released fixes for many of the reported issues.

The results suggest that specialized AI models could become increasingly valuable for finding vulnerabilities in widely used software, including applications that have already undergone extensive security testing.

However, the findings do not mean that every reported vulnerability has been fixed.

OpenAI Reports 948 Working Linux Exploits

GPT-6 Cyber was also used in research involving the Linux kernel.

OpenAI reported that the model:

Reproduced 1,724 vulnerabilities.
Developed working exploits for 948 vulnerabilities.
Helped researchers disclose 304 critical findings to Linux maintainers, together with candidate patches.

These figures come from OpenAI’s research and should not be interpreted as evidence that the vulnerabilities were exploited against real-world users.

The company says it is working with maintainers to address additional issues.

GPT-6 Cyber vs GPT-5.6 Cyber: How Much Better Is It?

OpenAI reports substantial improvements over its previous cybersecurity model.

Benchmark	GPT-5.6 Cyber	GPT-6 Cyber
ExploitGym success rate	32.1%	48.5%
Avoiding unauthorized targets without deployment safety prompts	94.49%	99.93%

On ExploitGym, which evaluates whether models can exploit specified vulnerabilities in a controlled environment, GPT-6 Cyber achieved a 48.5% success rate, compared with 32.1% for GPT-5.6 Cyber.

The newer model also performed better at avoiding systems outside its authorized testing scope.

OpenAI says GPT-6 Cyber stayed within the defined boundaries in 100% of one evaluation’s runs when deployment safety prompts were enabled.

These are controlled benchmark results, not a guarantee of perfect behavior in real-world environments.

Why Does OpenAI Classify GPT-6 Cyber as Critical?

OpenAI has classified GPT-6 Cyber at the Critical cybersecurity capability level under its Preparedness Framework.

Its predecessor, GPT-5.6 Cyber, was classified at the High level.

The Critical classification reflects the model’s advanced cybersecurity capabilities and the potential risks associated with misuse.

To reduce those risks, OpenAI says it has introduced additional safeguards, including monitoring for unauthorized activity, access restrictions and protections around critical infrastructure.

The company also recommends human oversight and isolated testing environments for high-risk security workflows.

Who Can Access GPT-6 Cyber?

GPT-6 Cyber is not generally available in ChatGPT.

At launch, access is limited to a selected group of approved customers in OpenAI’s Daybreak Red program.

Eligible customers can use the model through:

OpenAI’s API.
Codex, OpenAI’s coding environment.

Daybreak Red is intended for verified security teams performing advanced vulnerability research and authorized testing.

OpenAI says it plans to expand GPT-6 Cyber access to additional approved Daybreak Red customers.

However, the company has not announced a general public release date or standard consumer pricing.

Organizations interested in access can review the official OpenAI announcement for information about the Daybreak program.

What Does GPT-6 Cyber Mean for Cybersecurity?

GPT-6 Cyber highlights how quickly AI-assisted vulnerability research is advancing.

For defenders, more capable models could help identify software weaknesses earlier, validate security fixes and reduce the time required to investigate complex vulnerabilities.

But the same technical capabilities also raise concerns about misuse.

This creates a difficult challenge for AI developers: giving legitimate security researchers powerful tools while preventing unauthorized exploitation.

OpenAI’s restricted rollout suggests that access controls and safety monitoring will remain central to how advanced cybersecurity models are deployed.

For everyday ChatGPT users, the launch does not introduce a new general-purpose model or change existing ChatGPT subscriptions.

Readers interested in how OpenAI’s consumer AI compares with competing assistants can also explore our ChatGPT vs Claude comparison.

Frequently Asked Questions

Is GPT-6 Cyber available in ChatGPT?

No. GPT-6 Cyber is initially available only to selected approved Daybreak Red customers through the API and Codex.

Is GPT-6 Cyber better than GPT-5.6 Cyber?

In OpenAI’s published cybersecurity evaluations, GPT-6 Cyber outperformed GPT-5.6 Cyber in exploit development and staying within authorized testing boundaries.

Did GPT-6 Cyber hack Firefox?

OpenAI researchers used the model to identify 81 previously unknown high-severity Firefox vulnerabilities. The findings were disclosed to Mozilla, which has released fixes for many of them. This was authorized security research, not a reported attack on Firefox users.

How much does GPT-6 Cyber cost?

OpenAI has not announced standard public pricing for GPT-6 Cyber.

Final Thoughts

GPT-6 Cyber is a significant development in AI-powered cybersecurity, combining stronger vulnerability research capabilities with additional safeguards intended to limit unauthorized use.

The reported Firefox and Linux findings demonstrate the model’s potential value for security teams.

For now, however, GPT-6 Cyber remains a restricted research tool rather than a broadly available ChatGPT feature.

Sources

Editorial disclosure: This article is based on OpenAI’s November 5, 2026 announcement and independent news reporting. Performance figures and vulnerability counts are attributed to OpenAI and have not been independently reproduced by Radavora. The article does not constitute hands-on testing of GPT-6 Cyber.

OpenAI Launches GPT-6 Cyber, Finds 81 Firefox Security Flaws