Why a new security flaw in Java has triggered urgent advisories worldwide
What is Java4Shell? How did security researchers uncover it using AI? How is it different from the PHP2Shell vulnerability last month? And why are security agencies urging organisations to implement patches immediately?
Security researchers have found a critical vulnerability in Java, the popular open-source programming language used by several government agencies and most Fortune 500 companies.
The vulnerability called Java4Shell could enable hackers to gain remote access to sensitive internal systems. It can be exploited by having Java erroneously execute code and tampering with a section of computer memory, which could lead to the software programme crashing almost immediately.
Listed as ‘CVE-2026-114938’ in the global vulnerability database, Java4Shell was found to affect those using versions Java 17.0.6 on systems running x86-64 Intel and AMD chips with AVX-512 VBMI support. Java4Shell was first brought to light by researchers at cybersecurity startup Calif on Tuesday, November 17. They said that they were able to find the flaw using OpenAI’s recently launched GPT-6 Cyber AI model.
Oracle, owner of the Java trademark and provider of commercial Java licenses, released a software patch for the same on Tuesday. The OpenJDK project, in whose open-source code the bug was found, has also released a patch. Since the OpenJDK project underpins Java builds from many vendors, even those who do not use Oracle’s Java distributions, could be at risk, including enterprise applications, identity platforms, etc.
Amazon has issued fixes for its Corretto Java builds while Red Hat is also working on swiftly rolling out patches for its own Java builds. Other at-risk Java distributions include Eclipse Temurin, Microsoft Build of OpenJDK, Azul Zulu, BellSoft Liberica, and SAP SapMachine.
Java as a programming language is widely used by tech giants, banks, hospitals, airlines, and government agencies, including those with national security missions. This disclosure is the latest example of AI tools being used to uncover critical security vulnerabilities in widely used software, highlighting how much of the internet’s underlying infrastructure rests on systems with long-standing security flaws.
The same Calif researchers had found ways to exploit security flaws in PHP, an open-source programming language, last month. The PHP2Shell vulnerabilities exposed websites to remote code execution attacks, forcing site owners to urgently patch their systems to prevent hackers from taking control.
As access to increasingly capable AI models and agents become widespread, finding and potentially exploiting these zero-day flaws is likely to become faster and easier. To be sure, there is no evidence yet to show that Java4Shell has already been exploited by hackers to launch attacks.
The Java4Shell flaw was found in a performance shortcut that is used by Java to decode Base64 data faster on certain Intel and AMD processors. Base64 is a common algorithm that is used to translate binary computer data into legible text with lowercase and uppercase letters, numbers, punctuation, etc.
“When the shortcut receives a specially crafted Base64 string, it writes data beyond the memory it set aside. That can corrupt other information the application is holding in memory,” Satnam Narang, a researcher at cybersecurity firm Tenable, said.
There are two ways an attacker could launch an attack by exploiting Java4Shell: They can reach the application’s memory, which may contain credentials such as keys and tokens. The second method involves an attacker running their own malicious code on the server.
In its proof-of-concept attacks, Calif researchers said that they exploited Java4Shell to break into Keycloak, an authentication and login system used by thousands of organisations, including CERN, HP, UnitedHealthcare, and NASA.
Other applications that could be compromised by attackers without having accounts, are Atlassian Jira, Atlassian Confluence, JFrog Artifactory, Jenkins, PingFederate, Apache Kafka Broker, Apache Solr, Elasticsearch, Gerrit, and Spinnaker.
In some cases, the researchers found that when Java runs the vulnerable code, it tries to decode 64 “A” characters followed by one equal sign. Researchers were further able to demonstrate that they could gain full unauthorised access to the underlying system by using AI to corrupt data and carefully arrange the memory – a technique known as ‘heap grooming’
Calif’s Java research “shows how AI can assist expert security researchers with finding serious flaws in software we all depend on. We want more security teams to have these capabilities and tools to help make critical software more secure for everyone,” Fouad Matin, who leads cyber preparedness and safety at OpenAI, was quoted as saying on the use of OpenAI’s model to uncover the vulnerability.
The researchers’ findings were reportedly corroborated by other labs as well, who have been able to reproduce the crashes and a memory leak. They have said that they are holding back their product-specific exploits until organisations have had time to upgrade, and plan to publish them later.
Calif is also working with intelligence-sharing alliance Five Eyes — which consists of the United States, Canada, Australia, New Zealand and Britain — and major vendors in the private sector, including Google and Microsoft, according to a report by The New York Times.
“There is one limit worth knowing. No universal exploit exists: each product needs its own chain, and the technique is sensitive to network conditions. That slows mass exploitation. It does not protect a specific target such as your identity provider,” Saeed Abbasi, head of Threat Research Unit (TRU) and director of Product at cybersecurity firm Qualys, told The Indian Express.
Eduardo Vela Nava, an information security engineer at Google, said that “once notified, we swiftly rolled out mitigations and patches across Google services, developed detection to ensure there was no exploitation of the vulnerability, and provided guidance to our customers,” according to NYT.
In an advisory issued on Tuesday, the US Cybersecurity and Infrastructure Security Agency (CISA) urged organisations “to immediately patch the recently disclosed Oracle Java vulnerability.” “This critical memory-corruption vulnerability, known as “Java4Shell,” affects the Base64 decoder of the x86-64 HotSpot Java Virtual Machine (JVM), which is part of the Java Development Kit (JDK). Specially crafted Base64 input could crash the JVM and allow for unauthenticated remote code execution,” it said.
The Indian Computer Emergency Response Team (CERT-In) issued an advisory warning that Java4Shell could allow a remote attacker to execute arbitrary code on the targeted system, and advised those using affected versions of Oracle Java SE and Oracle GraalVM products to implement available patches immediately.
The Canadian Centre for Cyber Security also issued an alert stating that “organizations should identify affected systems and apply applicable vendor updates as soon as they become available.” “The Cyber Centre is not aware of active exploitation at this time, but that can change quickly,” it added.
On what should be the immediate steps, cybersecurity expert Abbasi said, “Reduce exposure while you wait for a patch. Take identity, build, and collaboration consoles off the open internet, or put them behind a VPN, until they are patched. CISA also suggests validating Base64 input and enforcing input-size limits at the application boundary. Size limits alone will not stop this bug, because the malformed input is very short. The JVM flag above is the more direct stopgap because it removes the vulnerable code.”
Last month, security researchers discovered flaws in PHP, an open-source programming language estimated to be used by 70 per cent of websites and apps on the internet.
The discovery was made by human researchers with the assistance of OpenAI models. It set off fears around the world about significant disruptions for consumers and businesses as companies raced to patch the vulnerability before hackers could exploit them.
The US National Security Agency (NSA) warned businesses to take down websites if they were not able to immediately deploy a patch. While the two critical bugs uncovered in PHP could have affected how websites function, the Java flaw has potentially wider implications as the programming language underpins cloud computing architecture, identity management tools, email services, and more.
It comes on the heels of another security incident, where some of the world’s most visited websites such as Zoom, HSBC, Equifax, CNN, Salesforce, and Ikea, were compromised by hackers using malicious JavaScript. The hacking campaign was aimed at stealing cryptocurrency and infecting iPhones, according to security researcher Scott Helme.
The malicious code was distributed to websites belonging to 178 organisations across multiple sectors through Optimizely, a website optimisation platform. This week, the US Federal Bureau of Investigation (FBI) also issued an alert warning that hackers were using new autonomous AI hacking tools such as ARTEX, to target government agencies and businesses around the world.
Security researchers have found a critical vulnerability in Java, the popular open-source programming language used by several government agencies and most Fortune 500 companies.
The vulnerability called Java4Shell could enable hackers to gain remote access to sensitive internal systems. It can be exploited by having Java erroneously execute code and tampering with a section of computer memory, which could lead to the software programme crashing almost immediately.
Listed as ‘CVE-2026-114938’ in the global vulnerability database, Java4Shell was found to affect those using versions Java 17.0.6 on systems running x86-64 Intel and AMD chips with AVX-512 VBMI support. Java4Shell was first brought to light by researchers at cybersecurity startup Calif on Tuesday, November 17. They said that they were able to find the flaw using OpenAI’s recently launched GPT-6 Cyber AI model.
Oracle, owner of the Java trademark and provider of commercial Java licenses, released a software patch for the same on Tuesday. The OpenJDK project, in whose open-source code the bug was found, has also released a patch. Since the OpenJDK project underpins Java builds from many vendors, even those who do not use Oracle’s Java distributions, could be at risk, including enterprise applications, identity platforms, etc.
Amazon has issued fixes for its Corretto Java builds while Red Hat is also working on swiftly rolling out patches for its own Java builds. Other at-risk Java distributions include Eclipse Temurin, Microsoft Build of OpenJDK, Azul Zulu, BellSoft Liberica, and SAP SapMachine.
Java as a programming language is widely used by tech giants, banks, hospitals, airlines, and government agencies, including those with national security missions. This disclosure is the latest example of AI tools being used to uncover critical security vulnerabilities in widely used software, highlighting how much of the internet’s underlying infrastructure rests on systems with long-standing security flaws.
The same Calif researchers had found ways to exploit security flaws in PHP, an open-source programming language, last month. The PHP2Shell vulnerabilities exposed websites to remote code execution attacks, forcing site owners to urgently patch their systems to prevent hackers from taking control.
As access to increasingly capable AI models and agents become widespread, finding and potentially exploiting these zero-day flaws is likely to become faster and easier. To be sure, there is no evidence yet to show that Java4Shell has already been exploited by hackers to launch attacks.
The Java4Shell flaw was found in a performance shortcut that is used by Java to decode Base64 data faster on certain Intel and AMD processors. Base64 is a common algorithm that is used to translate binary computer data into legible text with lowercase and uppercase letters, numbers, punctuation, etc.
“When the shortcut receives a specially crafted Base64 string, it writes data beyond the memory it set aside. That can corrupt other information the application is holding in memory,” Satnam Narang, a researcher at cybersecurity firm Tenable, said.
There are two ways an attacker could launch an attack by exploiting Java4Shell: They can reach the application’s memory, which may contain credentials such as keys and tokens. The second method involves an attacker running their own malicious code on the server.
In its proof-of-concept attacks, Calif researchers said that they exploited Java4Shell to break into Keycloak, an authentication and login system used by thousands of organisations, including CERN, HP, UnitedHealthcare, and NASA.
Other applications that could be compromised by attackers without having accounts, are Atlassian Jira, Atlassian Confluence, JFrog Artifactory, Jenkins, PingFederate, Apache Kafka Broker, Apache Solr, Elasticsearch, Gerrit, and Spinnaker.
In some cases, the researchers found that when Java runs the vulnerable code, it tries to decode 64 “A” characters followed by one equal sign. Researchers were further able to demonstrate that they could gain full unauthorised access to the underlying system by using AI to corrupt data and carefully arrange the memory – a technique known as ‘heap grooming’
Calif’s Java research “shows how AI can assist expert security researchers with finding serious flaws in software we all depend on. We want more security teams to have these capabilities and tools to help make critical software more secure for everyone,” Fouad Matin, who leads cyber preparedness and safety at OpenAI, was quoted as saying on the use of OpenAI’s model to uncover the vulnerability.
The researchers’ findings were reportedly corroborated by other labs as well, who have been able to reproduce the crashes and a memory leak. They have said that they are holding back their product-specific exploits until organisations have had time to upgrade, and plan to publish them later.
Calif is also working with intelligence-sharing alliance Five Eyes — which consists of the United States, Canada, Australia, New Zealand and Britain — and major vendors in the private sector, including Google and Microsoft, according to a report by The New York Times.
“There is one limit worth knowing. No universal exploit exists: each product needs its own chain, and the technique is sensitive to network conditions. That slows mass exploitation. It does not protect a specific target such as your identity provider,” Saeed Abbasi, head of Threat Research Unit (TRU) and director of Product at cybersecurity firm Qualys, told The Indian Express.
Eduardo Vela Nava, an information security engineer at Google, said that “once notified, we swiftly rolled out mitigations and patches across Google services, developed detection to ensure there was no exploitation of the vulnerability, and provided guidance to our customers,” according to NYT.
In an advisory issued on Tuesday, the US Cybersecurity and Infrastructure Security Agency (CISA) urged organisations “to immediately patch the recently disclosed Oracle Java vulnerability.” “This critical memory-corruption vulnerability, known as “Java4Shell,” affects the Base64 decoder of the x86-64 HotSpot Java Virtual Machine (JVM), which is part of the Java Development Kit (JDK). Specially crafted Base64 input could crash the JVM and allow for unauthenticated remote code execution,” it said.
The Indian Computer Emergency Response Team (CERT-In) issued an advisory warning that Java4Shell could allow a remote attacker to execute arbitrary code on the targeted system, and advised those using affected versions of Oracle Java SE and Oracle GraalVM products to implement available patches immediately.
The Canadian Centre for Cyber Security also issued an alert stating that “organizations should identify affected systems and apply applicable vendor updates as soon as they become available.” “The Cyber Centre is not aware of active exploitation at this time, but that can change quickly,” it added.
On what should be the immediate steps, cybersecurity expert Abbasi said, “Reduce exposure while you wait for a patch. Take identity, build, and collaboration consoles off the open internet, or put them behind a VPN, until they are patched. CISA also suggests validating Base64 input and enforcing input-size limits at the application boundary. Size limits alone will not stop this bug, because the malformed input is very short. The JVM flag above is the more direct stopgap because it removes the vulnerable code.”
Last month, security researchers discovered flaws in PHP, an open-source programming language estimated to be used by 70 per cent of websites and apps on the internet.
The discovery was made by human researchers with the assistance of OpenAI models. It set off fears around the world about significant disruptions for consumers and businesses as companies raced to patch the vulnerability before hackers could exploit them.
The US National Security Agency (NSA) warned businesses to take down websites if they were not able to immediately deploy a patch. While the two critical bugs uncovered in PHP could have affected how websites function, the Java flaw has potentially wider implications as the programming language underpins cloud computing architecture, identity management tools, email services, and more.
It comes on the heels of another security incident, where some of the world’s most visited websites such as Zoom, HSBC, Equifax, CNN, Salesforce, and Ikea, were compromised by hackers using malicious JavaScript. The hacking campaign was aimed at stealing cryptocurrency and infecting iPhones, according to security researcher Scott Helme.
The malicious code was distributed to websites belonging to 178 organisations across multiple sectors through Optimizely, a website optimisation platform. This week, the US Federal Bureau of Investigation (FBI) also issued an alert warning that hackers were using new autonomous AI hacking tools such as ARTEX, to target government agencies and businesses around the world.