OpenAI launches GPT-6 Cyber with ‘Critical’ cybersecurity rating

OpenAI has launched GPT-6 Cyber, a specialised artificial intelligence model designed to help cybersecurity researchers identify software vulnerabilities, develop and validate exploits, and fix security flaws before malicious actors can misuse them.

Advertisement

Announced on November 5, 2026, the model is initially available to a select group of approved customers. OpenAI has classified it as ‘Critical’ for cybersecurity capabilities under its Preparedness Framework, upgrading the rating from ‘High’ assigned to its earlier GPT-5.6 Cyber model.

The launch comes amid growing concerns about the misuse of advanced AI tools for cyberattacks. OpenAI said it has observed increased attempts by malicious cyber actors to misuse its models.

Identifies vulnerabilities in Firefox and Linux

OpenAI said its researchers have used GPT-6 Cyber to identify previously unknown security vulnerabilities in widely used software, including Mozilla Firefox and the Linux kernel.

Advertisement

According to the company’s findings:

• Mozilla Firefox: The model helped identify 81 previously unknown vulnerabilities, which Mozilla classified as high severity. OpenAI shared reproducible tests and proposed patches with Mozilla.
• Linux kernel: GPT-6 Cyber reproduced 1,724 vulnerabilities and developed working exploits for 948 of them.
• Security disclosures: OpenAI said it disclosed 304 critical findings to Linux maintainers, along with proposed patches.

OpenAI said Mozilla has started releasing fixes for many of the newly identified Firefox vulnerabilities. These figures are based on the company’s own research.

Also Explore: Applied AI for HR Professionals: Certification Program

Access restricted to approved cybersecurity teams

Advertisement

OpenAI is initially providing GPT-6 Cyber through its API and Codex to a select group of approved customers under its Daybreak Red programme.

The programme is intended for authorised cybersecurity professionals and organisations conducting advanced security work, including penetration testing, vulnerability research and exploit development.

OpenAI plans to gradually expand access to other approved Daybreak Red customers. It is also making its earlier GPT-5.6 Cyber model available to all customers under the programme.

The restricted rollout reflects the risks associated with advanced cybersecurity AI. While these models can help security teams identify and fix software weaknesses, their capabilities could also be misused to develop cyberattacks.

OpenAI said it has introduced additional safeguards, including cybersecurity-specific monitoring and measures to keep the model’s activities within the authorised scope of a research assignment.

What the launch means for cybersecurity

The launch highlights the growing role of AI in software security. Specialised models can help researchers examine large codebases, investigate vulnerabilities and develop potential fixes.

However, the technology also raises concerns about the misuse of AI for finding and exploiting security weaknesses. OpenAI’s Critical classification reflects the need for stronger access controls and safeguards as these capabilities advance.

The effectiveness of GPT-6 Cyber will depend on how reliably cybersecurity teams can use it to identify genuine vulnerabilities, verify findings and improve software security.