OpenAI Says Its AI Model Developed Exploits for 948 Linux Kernel Vulnerabilities

OpenAI has announced that its GPT-6 Cyber model successfully developed working exploits for 948 vulnerabilities affecting the Linux kernel. The achievement demonstrates the model’s ability to identify weaknesses in complex software systems and develop methods for exploiting them. The company disclosed the results as part of its broader announcement about the cybersecurity capabilities of its latest artificial intelligence technology.

According to OpenAI, the model’s capabilities extend beyond identifying security flaws to developing practical exploits that demonstrate how vulnerabilities could be abused. The Linux kernel, which forms the foundation of numerous operating systems and powers servers, computers and cloud infrastructure worldwide, is a particularly important target for cybersecurity researchers. Discovering weaknesses in the kernel before malicious hackers exploit them could help organisations strengthen their digital defences.

However, the findings also highlight growing concerns about the potential misuse of advanced AI systems. Technology capable of automatically discovering vulnerabilities and developing exploits could significantly reduce the technical expertise required to conduct sophisticated cyberattacks. OpenAI says it has introduced safeguards and restricted access to GPT-6 Cyber, initially making it available to approved cybersecurity professionals and organisations involved in authorised security research.

The development signals a major shift in the cybersecurity industry, where artificial intelligence is increasingly being used for both offensive and defensive operations. Security experts believe advanced AI models could help identify software weaknesses faster than traditional testing methods, allowing developers to release patches before vulnerabilities are exploited. However, the same capabilities could create serious risks if they become accessible to cybercriminals, making responsible deployment and stronger security controls increasingly important.