Friday, 20 November 2026

AI-Assisted Research Uncovers Critical Java4Shell Flaw in Just One Hour

Cybersecurity researchers used AI to discover a critical vulnerability in Java's memory handling in approximately one hour. Named Java4Shell, the flaw could allow attackers to compromise enterprise servers, prompting Oracle to release security patches.

Key points

  • Calif researcher Quang Luong discovered Java4Shell with AI assistance in around one hour, with an initial exploit developed two hours later.
  • The vulnerability could enable remote code execution in affected Java applications, including Jenkins, Jira, Confluence and Keycloak.
  • Oracle and OpenJDK released security fixes on 17 November 2026, while CISA urged organisations to patch immediately.

A cybersecurity researcher has used artificial intelligence to uncover a serious vulnerability in Java, one of the world’s most widely deployed software platforms, in approximately one hour.

The vulnerability, named Java4Shell and tracked as CVE-2026-114938, affects OpenJDK’s Base64 decoder and could allow attackers to access sensitive information or remotely execute malicious code on vulnerable systems.

Security firm Calif disclosed the flaw on 17 November 2026, revealing that researcher Quang Luong identified it with the assistance of OpenAI’s Codex technology. The researchers reportedly used OpenAI’s GPT-6 Cyber model, which is not publicly available and only used “by banks, hospitals, the aviation industry and technology giants, as well as government agencies, including those with national security missions.”

According to Calif, Luong discovered the vulnerability in about an hour and developed an initial working exploit within another two hours. The firm said the AI token costs amounted to less than $200.

The discovery illustrates how AI-assisted security research can accelerate the identification and exploitation of vulnerabilities in widely used software.

A simple string exposes a serious Java vulnerability

Java4Shell originates from a memory-corruption error in OpenJDK’s hardware-accelerated Base64 decoder.

Base64 is a commonly used encoding format for transferring information such as authentication credentials and application data.

On affected systems, a specially crafted input containing 64 consecutive capital ‘A’ characters followed by an equals sign can trigger the vulnerability. Some older OpenJDK 17 builds require a longer variation of the input.

The malformed data causes the decoder to write beyond its allocated memory buffer before Java rejects the input.

Ordinarily, such an error could crash the Java Virtual Machine (JVM). However, Calif demonstrated that carefully manipulating memory allocation can turn the flaw into a more serious attack, potentially allowing unauthorised access to sensitive data and remote code execution.

The researchers rated Java4Shell 9.2 out of 10 under CVSS 4.0, while Oracle assigned it a high-severity score of 8.1 under CVSS 3.1.

Major enterprise applications could be exposed

Calif demonstrated remote code execution without authentication against default configurations of several widely used Java-based applications, including Keycloak, Atlassian Jira, Confluence, Jenkins, JFrog Artifactory and PingFederate.

The researchers also demonstrated attacks requiring authentication against Apache Kafka Broker, Apache Solr, Elasticsearch, Gerrit and Netflix Spinnaker.

However, not every Java installation is vulnerable.

The flaw affects specific OpenJDK versions running on x86-64 processors that support AVX-512 VBMI instructions. ARM64 systems and processors without the required instruction support are not affected by this particular vulnerability.

Calif also cautioned that there is no universal exploit capable of compromising every affected application. Successful attacks require techniques adapted to the targeted software and its memory behaviour.

Oracle releases fixes as security agencies issue warnings

Oracle and the OpenJDK project released security fixes on 17 November 2026, shortly after the vulnerability was publicly disclosed.

The US Cybersecurity and Infrastructure Security Agency (CISA) subsequently urged organisations to identify affected Java installations and apply the available updates immediately.

Administrators are advised to check Java runtimes bundled with third-party applications, container images and enterprise platforms, as updating the system-wide Java installation may not protect software using its own runtime.

Where immediate patching is not possible, Calif recommends temporarily disabling the vulnerable hardware-accelerated Base64 implementation using the JVM option -XX:-UseBASE64Intrinsics.

The researchers have demonstrated working exploits, but successful exploitation in real-world attacks had not been confirmed in the security advisories available at the time of reporting.

AI accelerates the search for software vulnerabilities

The involvement of AI is particularly notable because Java has been extensively examined by security researchers over many years.

Calif explained that Luong directed Codex towards complex areas of OpenJDK, including its just-in-time compiler, processor-specific optimisations and widely accessible APIs.

After several unsuccessful findings, the AI-assisted investigation identified the Base64 memory-corruption issue.

The company subsequently used AI agents to accelerate the development of application-specific exploitation techniques, although human researchers remained involved in directing and validating the work.

The case demonstrates both the defensive value of AI-assisted vulnerability research and the possibility that similar tools could help attackers develop exploits more quickly.

For organisations relying on Java, the immediate priority remains identifying vulnerable runtimes and deploying the available security updates.

KA

Founder and editor of Tech Prolonged since 2010. A software engineer by profession, Khurram covers Google, Android and Microsoft, and edits the newsroom's coverage. Find him on X at @khurramar.