Wednesday, 18 November 2026NewsWorldBusinessTech
Latest

Calif Finds Java4Shell Flaw Threatening Global Digital Infrastructure

The cybersecurity firm Calif discovered a critical security flaw in the Java programming language, called Java4Shell, using OpenAI’s GPT-6 Cyber model on Tuesday.

Remote attackers can execute unauthorized commands by corrupting memory within the OpenJDK Base64 decoder via vulnerability CVE-2026-114938.

Java powers about 90 percent of Fortune 500 companies in banking, aviation, and telecommunications. This flaw puts global digital infrastructure at risk, according to reports from BBC and infobae.

The flaw targets the hardware-accelerated Base64 decoding process on x86-64 processors. An attacker can trigger a calculation error by providing an input of 64 characters followed by an equals sign, El Ecosistema Startup reported. This error results in a negative output length, causing the system to write data beyond the intended memory buffer.

Technical analysis via Phoronix explained that the overflow occurs during processing, even though the system eventually rejects the input. The routine may alter neighboring objects, which could corrupt memory by as much as 62 bytes beyond its limit.

Severity ratings differ by scale

Oracle cataloged the vulnerability with a CVSS 3.1 score of 8.1 and classified it as high severity. But Calif assigned it a critical score of 9.2 under the CVSS 4.0 scale, RedesZone said.

Calif Finds Java4Shell Flaw Threatening Global Digital Infrastructure
Photo: Infobae

Exploiting the vulnerability depends on specific hardware and software. It requires x86-64 processors that are compatible with AVX-512 VBMI instructions. These are common in modern data centers, DiarioBitcoin reported.

Affected x86 families include Intel Xeon processors from the Ice Lake generation. This includes Sapphire Rapids, Emerald Rapids, and Granite Rapids. Affected groups also include AMD EPYC and Ryzen processors from the Zen 4 architecture, El Ecosistema Startup said.

Devices using ARM64 processors are not affected, infobae said.

The impact depends on application security settings. In some environments, an attacker can achieve remote code execution without any prior authentication.

Keycloak, Atlassian Jira, Atlassian Confluence, JFrog Artifactory, Jenkins, and PingFederate allow unauthenticated execution, RedesZone said.

Attackers may need existing credentials for other services. This applies to Apache Kafka Broker, Apache Solr, Elasticsearch, Gerrit, and Spinnaker, RedesZone added.

Calif released a proof of concept to show memory corruption. The firm did not release functional exploits for Jenkins or Jira to avoid helping attacks while organizations patch.

The following table compares the different severity ratings assigned to the vulnerability:

Rating Source CVSS Version Score
Calif 4.0 9.2
Oracle 3.1 8.1

Oracle issues emergency patches

Oracle released a security update on November 17 to address Java4Shell. This update is part of a November critical security patch that corrects 655 vulnerabilities across Oracle’s product catalog, RedesZone reported.

Oracle identifies 17.0.21.0.3 as affected for Java SE 17. Datadog lists the corrected builds as 17.0.21.1+4, 21.0.13.1+4, 25.0.5.1+4, and 27.0.1.1+4.

Aikido reports the corrected builds as 17.0.21.1, 21.0.13.1, 25.0.5.1, and 27.0.1.1.

The vulnerability affects various Java distributions built on OpenJDK code. Exposed versions include Oracle Java SE, Amazon Corretto, Eclipse Temurin, Microsoft Build of OpenJDK, Azul Zulu, and Red Hat OpenJDK, RedesZone said.

Red Hat said it is accelerating the preparation of its own patches. These patches were not yet published at the time of reporting.

The United States Cybersecurity and Infrastructure Security Agency (CISA) urged organizations to apply the patch immediately. CISA added the vulnerability to its catalog of exploited vulnerabilities, CriptoNoticias reported.

U.S. federal institutions must correct the flaw by December 1, CriptoNoticias said.

Administrators can temporarily disable hardware-accelerated Base64 decoding to reduce risk before an update. They can start OpenJDK with the parameter -XX:-UseBASE64Intrinsics, RedesZone said. They can also disable AVX-512 instructions using -XX:UseAVX=2 on compatible systems.

The Canadian Centre for Cyber Security recommends prioritizing internet-facing services. They also recommend verifying the specific Java provider and build used by every application.

AI speeds up vulnerability detection

The discovery shows how artificial intelligence can find security flaws. Greg Brockman, president of OpenAI, linked the discovery to the GPT-6 Cyber model, Fomoera reported. The model was recently released with limited access to approved organizations.

Calif coordinated with intelligence partners of the Five Eyes alliance before making the information public. This alliance includes the United States, Canada, the United Kingdom, Australia, and New Zealand. Calif also coordinated with major cloud providers like Google and Microsoft to allow defenders to prepare patches before criminals, Euronews said.

The discovery of Java4Shell is the second major flaw Calif has identified recently using OpenAI models, following the discovery of vulnerabilities in PHP. The discovery of Java4Shell shows how AI is accelerating the detection of weaknesses in widely used programs, The New York Times reported.

“this is a flaw so improbable and grave,” said Toby Walsh, a prominent artificial intelligence researcher.