An AI model has uncovered a critical vulnerability in Java itself, the programming language used by roughly 90% of Fortune 500 companies, and the flaw can be exploited to take complete control of affected systems.
Security startup Calif used OpenAI's GPT-6 Cyber, released earlier this month, to find the bug, dubbed Java4Shell, buried in the Java routine that converts binary to text. A trigger string of 64 capital A's followed by an equals sign crashes the program, and paired with a 2007 technique known as heap grooming, an attacker can hijack the entire system.
Oracle, which owns Java, has acknowledged the flaw and released a patch, while CISA issued an urgent advisory telling organizations to audit software inventories for JDKs and Java runtimes, including those bundled with third-party applications or packed into container images and appliances.
Calif CEO Thai Duong said turning the flaw into a working attack would not be easy, buying administrators time to patch on a non-emergency basis, but warned that vulnerabilities have a very long tail. Last month the company publicized a separate PHP flaw, found with the earlier GPT-5.6 Sol model, that likely affects millions of websites.