OpenAI Launches Cybersecurity-Focused AI Model Amid Rising Digital Threats

Image: TechCrunch · Source

OpenAI introduces ChatGPT Cyber, a new AI model designed to enhance cybersecurity defenses by discovering and patching software vulnerabilities, responding to growing concerns about AI-aided hacking.

OpenAI has unveiled a specialized cybersecurity AI model named ChatGPT Cyber, aimed at identifying and addressing software vulnerabilities to protect critical digital infrastructure. This release comes as AI-powered tools increasingly aid both cybersecurity defenders and potential attackers.

The significance of such models was highlighted by Max Justicz, CTO and co-founder of Charge Robotics, who used a previous OpenAI model, GPT 5.6 Sol, to uncover a severe vulnerability in Ubuntu’s widely used Linux-based operating system. By instructing the AI to find bugs comparable in severity to a previously patched issue, the model discovered an intricate exploit that could allow hackers to take control of servers if exploited. This finding was reported to the Debian Project, which promptly issued a patch; there is no evidence the vulnerability has been exploited in the wild.

OpenAI’s newly released model is accessible to top-tier participants in the company’s Daybreak cybersecurity program, specifically within Daybreak Red, which focuses on offensive security research and vulnerability hunting. OpenAI revealed that this model recently scanned the Linux kernel and identified 948 potential vulnerabilities, alongside uncovering 81 new vulnerabilities in Firefox, demonstrating the AI’s capability to significantly enhance security auditing.

To bolster industry defense efforts, OpenAI has announced a $1 billion initiative supporting cybersecurity professionals working on critical infrastructure sectors such as electrical grids and water systems. However, the program has drawn some criticism regarding restricted access to powerful features, which some experts feel limits defenders’ effectiveness compared to attackers who might utilize unrestrained open-weight AI models.

Despite participation from thousands of cybersecurity experts, the scale is dwarfed by the vast number of potential malicious actors. OpenAI acknowledges this imbalance and is dedicating a substantial portion of its advanced computing resources to pro bono vulnerability discovery and patching in open-source codebases. Nevertheless, the cost of these efforts, such as the $1.2 million Linux bug hunt, is modest compared to other large-scale AI projects.

The emergence of ChatGPT Cyber underscores the pressing need for the cybersecurity community to leverage frontier AI technology promptly. As AI models advance, the risk window narrows, making coordinated efforts to identify and fix vulnerabilities increasingly urgent to safeguard the digital infrastructure underpinning modern society.