OpenAI's GPT-6 Cyber Finds Firefox, Linux Flaws; The Growing Stakes of AI-Powered Exploit Research

OpenAI has released GPT-6 Cyber, a new AI model designed to help cybersecurity experts identify software vulnerabilities and assess how attackers might exploit them. Announced on November 5, 2026, the model has received a “Critical” cybersecurity rating under OpenAI’s Preparedness Framework, a higher classification than the “High” rating assigned to its predecessor, GPT-5.6 Cyber. OpenAI said researchers had already used GPT-6 Cyber to discover and verify security weaknesses in widely used software, including Mozilla Firefox and the Linux operating system.

The company said the model is intended for authorised security research, allowing professionals to identify vulnerabilities, test potential exploits and develop solutions before malicious actors can take advantage of the weaknesses.

GPT-6 Cyber Finds Security Weaknesses in Firefox and Linux

One of the key findings shared by OpenAI concerns Mozilla Firefox, a widely used web browser. The company said its researchers used GPT-6 Cyber to identify previously unknown vulnerabilities in Firefox. Mozilla classified 81 of these as high-severity and began working on fixes. 

"During a vulnerability research campaign, GPT‑6 Cyber produced evidence and reproducible tests for 81 novel vulnerabilities classified by Mozilla as high severity, and showed that an additional 30 known issues were more severe than previously reported. We disclosed the findings to Mozilla, along with proposed patches to help its engineers develop fixes. Mozilla was supportive of this work, and began landing fixes shortly after receiving the reports⁠(opens in a new window)," OpenAI in a public statement. 

The model also helped researchers develop working exploits for 948 Linux kernel vulnerabilities, 588 of which were previously unknown, according to OpenAI. The company further said it had shared 304 critical findings with Linux maintainers and proposed patches to address the security issues. The Linux kernel is the core component of the Linux operating system. It manages communication between software and computer hardware, making its security critical for computers, servers and other digital infrastructure.

The statement added, "Developing exploits for 948 Linux kernel vulnerabilities⁠. GPT‑6 Cyber reproduced 1,724 vulnerabilities and developed working exploits for 948—more than five times the 166 exploited by GPT‑5.6 Sol using the same research harness. Following the Linux community's guidance, we disclosed 304 critical findings to maintainers with candidate patches and are in the process of getting other vulnerabilities fixed."

Why Has OpenAI Given It a ‘Critical’ Rating?

The “Critical” classification reflects the model’s advanced ability to identify vulnerabilities and demonstrate how they could be exploited. GPT-6 Cyber goes beyond identifying a software weakness. It can also assess whether the vulnerability could be used to gain unauthorised access to a system, allowing researchers to understand the potential consequences of a security flaw. OpenAI said the model outperformed its predecessor on several cybersecurity evaluations.

On ExploitGym, a benchmark that assesses a model’s ability to exploit specific vulnerabilities in controlled environments, GPT-6 Cyber scored 48.5%, compared with 32.1% for GPT-5.6 Cyber. However, these capabilities also raise concerns about potential misuse. A model that helps security researchers identify and test vulnerabilities could also assist malicious actors in discovering weaknesses and developing attacks. This makes access controls and the model’s ability to remain within the boundaries of authorised security testing particularly important.

Who Can Access GPT-6 Cyber?

OpenAI is not making GPT-6 Cyber publicly available at launch. The model is initially being offered to selected customers through Daybreak Red, the company’s programme for advanced and authorised cybersecurity work, including penetration testing, vulnerability research and exploit development. Penetration testing involves examining a computer system or application for security weaknesses through controlled and authorised tests.

OpenAI said it would extend access to all approved Daybreak Red customers over the next few weeks. The company is also making GPT-5.6 Cyber available to all Daybreak Red customers. The restricted rollout reflects the risks associated with a model capable of identifying vulnerabilities and demonstrating how they could be exploited.

What Does This Mean for Cybersecurity?

The launch highlights the growing role of artificial intelligence in protecting computer systems and digital infrastructure. Cybersecurity researchers often have to examine vast amounts of code to identify potential weaknesses. AI models such as GPT-6 Cyber could accelerate this process by detecting vulnerabilities, testing possible attack methods and suggesting fixes. However, the same capabilities could also make it easier for malicious actors to discover vulnerabilities and develop attacks.

OpenAI’s decision to restrict access highlights a central challenge facing AI companies: how to provide cybersecurity professionals with increasingly powerful tools while preventing those tools from being used against businesses, governments and critical services. As AI systems become more capable, determining who can access them, how they can be used and whether they remain within authorised limits is likely to become an increasingly important part of cybersecurity.