A serious security hole in Java, the software platform that sits underneath a huge share of the world's business applications, has sent IT teams scrambling to patch. The flaw, tracked as CVE-2026-114938 and nicknamed "Java4Shell" by the researchers who disclosed it, was made public on Tuesday, 17 November 2026. On the same day, Oracle shipped fixed versions and the US Cybersecurity and Infrastructure Security Agency (CISA) published an alert telling organisations to patch immediately.

What has drawn even more attention than the bug is how it was found. The security research firm Calif says artificial intelligence tools did much of the heavy lifting, both in spotting the flaw and in building attacks around it.

What happened

According to CISA's alert, the problem is a memory-corruption bug in the Base64 decoder of the x86-64 HotSpot Java Virtual Machine, which is part of the Java Development Kit (JDK). Base64 is a common way of converting data into plain text so that it can travel through web requests, emails and configuration files. CISA said specially crafted Base64 input can crash the Java Virtual Machine and allow an attacker to run code remotely without logging in.

The Hacker News explained that a malformed Base64 string ending in an extra "=" sign causes the decoder to write data beyond the space reserved for it, and Java rejects the string only after that write has already happened. CyberScoop, describing Calif's findings, said such an input can let an unauthenticated attacker corrupt server memory, read nearby data and potentially escalate to full control of the machine.

Who is exposed

The bug does not affect every computer running Java. Security firm Tenable, in an FAQ published on 17 November, said exposure requires an x86-64 processor that supports a set of instructions called AVX-512 VBMI, found in several recent Intel Xeon server chips and in AMD Zen 4 or newer processors. Calif says ARM64 systems are not affected. The application must also pass untrusted data to the affected Base64 functions.

On versions, Tenable said the OpenJDK Java 17, 21, 25 and 27 lines are affected, with Calif putting the starting point at OpenJDK 17.0.6. Oracle also listed certain Java SE Enterprise Performance Pack and GraalVM builds. According to Tenable and The Hacker News, standard Java 8 and Java 11 do not contain the vulnerable code.

Calif says that in its own tests, attacks worked without any login under default settings on Keycloak, Atlassian Jira, Atlassian Confluence, JFrog Artifactory, Jenkins and PingFederate. With a valid login, it says attacks worked on Apache Kafka, Apache Solr, Elasticsearch, Gerrit and Spinnaker. These are widely used tools for logins, project tracking, documentation, software builds and data search. Calif has said the list is not complete.

The AI angle

CyberScoop reported that Calif researcher Quang Luong found the bug in about an hour while auditing OpenJDK with the help of an OpenAI tool, at a cost of under $200 in computing charges. Calif said AI agents produced most of its exploit chains almost on their own, and described what it called the era of "the centaur hacker, part human, part AI". CyberScoop said this was Calif's second major AI-assisted discovery in recent weeks.

Gal Elbaz, co-founder and chief technology officer of Oligo Security, told CyberScoop that this was a vulnerability in one of the most widely used software runtimes in the world, "with initial exploit development taking just a few hours".

Timeline

  • 9 September 2026: According to The Hacker News, a malformed input that triggers the overflow was posted publicly in an OpenJDK pull request proposing a fix for the same decoder code. The request was closed without being merged in early November after an automated inactivity notice.
  • 17 November: Calif said it had found and reported the flaw and that details had leaked before a fix, Tenable reported. Oracle released its November Critical Security Patch Update and the OpenJDK Vulnerability Group published an advisory. Calif released a technical write-up and a limited proof-of-concept. CISA issued its alert late in the day.
  • 18 November: Fixes from other Java distributors were still arriving at different speeds.

Verified data

  • Fixed Oracle builds: Java 17.0.21.1, 21.0.13.1, 25.0.5.1 and 27.0.1.1, plus 8u513-perf for the Java 8 Enterprise Performance Pack, according to Tenable.
  • Oracle's November update covers 652 flaws across its products, CyberScoop reported.
  • Calif rates the bug 9.2 out of 10 under CVSS 4.0; Oracle rates it 8.1 under CVSS 3.1 and describes it as hard to exploit but needing no login.
  • Red Hat gave a preliminary score of 9.8 and Amazon scored it 8.8, The Hacker News reported.

Different perspectives

The researchers and the vendor do not fully agree on how dangerous the flaw is. Oracle's lower score reflects its view that attacks are complex to pull off. Its CVE record nonetheless warns that successful attacks "can result in takeover of Oracle Java SE". Amazon's description is more cautious, speaking of a crash for unauthenticated users and code execution in certain application environments.

Independent labs have so far confirmed only part of the picture. The Hacker News reported that other teams have reproduced crashes and a memory leak but not remote code execution. Upwind said it reproduced memory corruption on a cloud server with an Intel Ice Lake chip, while Blackpoint Cyber said it crashed a test application with about 28,300 malformed login requests in roughly 20 seconds but did not achieve code execution. Researchers nonetheless told CyberScoop the likelihood of real-world exploitation is high.

What is established

  • CVE-2026-114938 is a memory-corruption flaw in Java's Base64 decoder on x86-64 systems; Oracle, OpenJDK and CISA have all acknowledged it.
  • Oracle released fixed builds on 17 November, and CISA has urged immediate patching, starting with internet-facing systems.
  • No exploitation in the wild had been reported as of 17 November, and the flaw was not in CISA's Known Exploited Vulnerabilities catalogue. Canada's Cyber Centre said it was not aware of attacks.

What is uncertain

  • Real severity: scores range from Oracle's 8.1 to Red Hat's preliminary 9.8, and independent labs have not yet reproduced remote code execution.
  • Which AI tool was used: CyberScoop reported that Calif used OpenAI's Codex, while Tech Times, citing the Financial Times, said the model was OpenAI's GPT-6 Cyber.
  • The leak: it is not clear how widely the September pull request was noticed, or whether anyone used it before the fix.
  • Links to recent outages: Just Security's daily news roundup noted that officials have dismissed suggestions that recent outages, including one at the NSA, were linked to the flaw, attributing them instead to rushed patching. No source reviewed for this article has connected the bug to any outage.
  • Downstream fixes: Amazon Corretto, Azul Zulu and SapMachine had released fixed builds, but The Hacker News found no fixed builds listed for Eclipse Temurin, Microsoft Build of OpenJDK or BellSoft Liberica, and Red Hat had yet to ship fixes for its enterprise Linux versions.

What it means for Indian readers

For ordinary smartphone and laptop users, there is little to do beyond installing updates when prompted; the risk centres on servers that run Java applications. For Indian companies, banks, start-ups, colleges and government offices that run tools such as Jira, Confluence, Jenkins, Keycloak or Elasticsearch on recent Intel or AMD servers, the same exposure conditions described by Tenable apply. Tech Times noted that the concern extends to Java software used by banks and government agencies, though it reported no confirmed attack on those sectors. Reuters, as cited by Tech Times, reported that South Korea's presidential office held an emergency meeting with relevant agencies, a sign of how seriously some governments are treating it.

CISA's advice is a practical checklist for any IT team. It asks organisations to inventory every Java installation, including runtimes bundled inside third-party applications, container images and appliances, which may not appear separately in software lists. It then says to patch, prioritising internet-facing systems, and to watch for unexpected crashes or suspicious Java processes. Where patching is not immediately possible, CISA suggests validating and restricting Base64 input, enforcing size limits and limiting unauthenticated access. Calif has also suggested startup settings that switch off the vulnerable code path, though Tenable advises confirming they work on a given platform first.

What happens next

Calif is holding back application-specific attack code and says it will publish it later, according to Tenable, which means the window for patching before detailed exploits circulate may be short. Distributors that have not yet shipped fixes are expected to release them on their own schedules, and vendors of the affected applications may issue their own guidance. Security teams will also be watching whether CISA adds the flaw to its Known Exploited Vulnerabilities catalogue, which would signal confirmed attacks.

PS · Hugovanmeijeren, CC BY-SA 3.0