64 'A' Letters Threaten Global Giants: GPT-6 Finds Critical Vulnerability in Cyber Java
· AI Applications · Cem Koyluoglu
The cybersecurity startup Calif, which has been detecting security vulnerabilities that can be exploited in software one after another for the past few weeks, is known as “vulnopocalypse.”
TL;DR Calif, a cybersecurity startup, used OpenAI’s GPT‑6 Cyber to uncover a critical Java flaw dubbed Java4Shell, which could allow attackers to hijack systems via a specific string of 64 “A”s followed by “=”. Oracle has patched the issue and CISA issued an emergency update, urging users to check JDKs and runtimes. The discovery highlights the growing threat landscape and Calif’s role in exposing vulnerabilities.
Key Highlights • Calif used GPT‑6 Cyber to discover Java4Shell, a flaw in Java's binary‑to‑text routine. • Java4Shell can be triggered by 64 “A” characters followed by “=” and combined with heap grooming to take control of systems. • Oracle and CISA have released patches and emergency updates, urging users to audit JDKs and runtimes.
New Java Vulnerability Discovered by Calif Using GPT-6 Cyber
A cybersecurity startup named Calif has identified a critical flaw in the Java programming language, positioning itself as a leading actor in what the New York Times has dubbed the “vulnopocalypse.” According to the Times, Calif leveraged OpenAI’s newly released GPT‑6 Cyber to uncover a serious bug in Java, marking the company’s most striking achievement to date.
The flaw does not involve JavaScript used by a particular site or business; it resides directly within Java itself. This discovery is likened to a locksmith finding a hidden weakness in a door hinge that no one else has noticed. While the source of the flaw remains unclear, the Times noted that roughly 90 % of Fortune 500 companies use Java, underscoring the potential impact of the vulnerability.
Java4Shell
The vulnerability, referred to as Java4Shell, lies in Java’s routine that converts binary data to text. It can be triggered by a specific character sequence that resembles a cat walking across a keyboard: 64 consecutive “A” characters followed by an equals sign, represented as “AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA=”.
Independent testing by the Times confirmed that Calif researchers, using GPT‑6 Cyber, demonstrated how this seemingly innocuous string could transform into a security nightmare. An attacker can combine Java4Shell with the “heap grooming” or “heap feng shui” technique first identified in 2007 to seize control of the affected system.
Oracle, the owner of the Java brand, acknowledged the issue and has already released a patch. The U.S. Cybersecurity and Infrastructure Security Agency (CISA) has also issued an emergency update. CISA’s advisory urges users to review their software inventories for JDKs and Java runtime environments, including third‑party applications and container images.
CEO’s Perspective
Thai Duong, CEO of Calif, emphasized that exploiting the vulnerability is not straightforward. He suggested that the discovery provides administrators with a window to apply updates before the flaw can be weaponized. “However, the effects of security vulnerabilities can persist for a long time,” Duong told the Times.
Previous Work
Last month, Calif announced a PHP vulnerability that could potentially affect millions of websites, discovered using GPT‑5.6 Sol. This latest finding, identified by researcher Bruce Dang, targets a system he described as the “modern economy’s beating heart.”