OpenAI's GPT-6 Cyber Finds Thousands of Vulnerabilities

OpenAI has unveiled the specialized GPT-6 Cyber model, designed for penetration testing and cybersecurity. This neural network has demonstrated outstanding results in identifying vulnerabilities within complex software.

The model successfully breached the Linux operating system kernel, uncovering over a thousand distinct vulnerabilities. Specifically, 81 new critical security flaws were identified in the Firefox browser.

GPT-6 Cyber demonstrates effectiveness surpassing that of traditional hacking teams. The neural network conducts successful attacks on 95% of test servers, marking a 16% improvement over previous generations of cybersecurity models.

A key feature of this development is its built-in risk control system. When performing high-risk operations, the model requires mandatory approval from cybersecurity specialists and operates strictly within a secure perimeter.