
Java4Shell explained: affected Java versions and fixes

On November 17, 2026, Calif disclosed a memory corruption flaw in OpenJDK’s Base64 decoder. A crafted 65-character Base64 string can allow an attacker to run code on a Java-based system without logging in. The exploit has been nicknamed Java4Shell, a reference to 2021’s Log4Shell incident.
What is Java4Shell?
The vulnerability is a memory-corruption flaw in the Base64 decoder of the x86-64 HotSpot JVM. A maliciously crafted Base64 input can crash the JVM and may allow unauthenticated remote code execution. In Calif’s testing, it was able to achieve unauthenticated RCE in default configurations of popular tools like Atlassian Jira, Atlassian Confluence, and Jenkins.
The exploit affects Java 17.0.6 and later, 21, 25, and 27 (plus Oracle's Java SE 8 Performance Pack build 8u511b33-perf). The advisory does not list Standard Java 8 and 11 as affected.
The risk is highest for Java services that accept untrusted input (for example web services, APIs, and message consumers) and run on x86-64 hosts with AVX-512 VBMI support. Because the JRE is often baked into container images, affected runtimes may also be present in Docker images.
Oracle is strongly advising impacted users to upgrade to a patched Java release as soon as possible. Oracle's fixed builds are 17.0.21.1, 21.0.13.1, 25.0.5.1 and 27.0.1.1. Users of other OpenJDK distributions (Temurin, Corretto, Debian/Ubuntu/RHEL packages, etc) should check with the vendors, as fixed versions are still rolling out at the time of writing. Calif has also flagged mitigations that can be used while fixes are prepared.
AI is finding bugs that sat in code for years
Java4Shell is another example of AI tools uncovering bugs and potential exploits that sat undiscovered in commonly used services for years. In the case of CVE-2026-114938, the flawed assembly decoder was added to mainline OpenJDK in 2021. A Calif researcher identified the bug using AI tools in late October 2026.
Node2Shell sat in Google’s Brotli decompression library for years before it was identified, and PHP2Shell traces back to a class of bug that was first flagged in 2004.
The vast majority of researchers are following responsible disclosure practices. This includes sharing insights within the industry through coalitions like Athena, and waiting until fixes are in place before disclosing more widely. However, in the case of Java4Shell, details of the vulnerability leaked before fixes were in place.
The speed at which long-standing issues are being identified and exploited is putting pressure on teams managing on-prem infrastructure. The Zero Day Clock Observatory has tracked a 362% jump in CVE records year-over-year. Microsoft is already warning on-prem customers to expect a significant jump in vulnerability fixes. Small teams running on-prem instances risk being swamped by the sheer volume of vulnerabilities and the incredibly limited time available to patch them before exploits begin circulating.
Coordinated disclosure held up in part this time: Calif published its full write-up only after Oracle’s fixes shipped. But leaks are becoming more common, and the gap between a fix and a working exploit keeps shrinking.The key question for the industry now is whether a disclosure model built for human-paced research can keep up with AI-paced discovery.
More articles


Software supply chain defense in depth: A stage-by-stage guide

Critical PHP vulnerabilities in PHP2Shell expose web infrastructure to remote code execution

LiteLLM 1.107.0 compromised on PyPI

Why enforce supply chain security at the artifact registry

