Max Justicz, Chief Technology Officer at Charge Robotics, said he used a specialized OpenAI model to uncover a security vulnerability in widely used open-source software. The flaw could allow attackers to break into large data centers.
Justicz said he reported the vulnerability in late September to Canonical, the British company that maintains the Ubuntu operating system. The company plans to release a patch fixing the flaw on Wednesday.
This marks the latest case in recent months of researchers using advanced artificial intelligence models to find security defects in long-used software code. Discoveries have especially proliferated since OpenAI launched its bug-fixing initiative "Patch the Planet" in June and Anthropic introduced its similar "Cyber Mission" program in October.
Justicz said cybersecurity research is his hobby. Using the OpenAI GPT-5.6 Sol model, which is tailored for cybersecurity researchers, he found a flaw in apt, the system within the Ubuntu operating system that handles software downloads and package management.
He warned that malicious attackers could exploit the flaw by tapping into fiber-optic lines connected to data centers, running their own code on data center servers and hijacking websites.
Justicz said he has not yet found anyone attempting to exploit the vulnerability in an attack.
The apt flaw is the latest in a string of major cybersecurity vulnerabilities uncovered recently. Researchers previously found a flaw in mainstream website hosting software that affects 70% of websites worldwide, and Anthropic researchers also discovered security defects in the Linux code used by Ubuntu.
Comments