Security researchers have disclosed a memory-corruption vulnerability in OpenJDK that enables information disclosure and remote code execution across popular enterprise applications. Dubbed Java4Shell and tracked as CVE-2026-114938, the flaw was patched by Oracle on November 17 in a broader security update. Calif rates the severity at 9.2/10, while Oracle gave it an 8.1 score. The issue impacts OpenJDK versions 17.0.6 and later, alongside distributions such as Amazon Corretto, Microsoft Build of OpenJDK, and Red Hat OpenJDK. [1][3][4]
The bug resides in hardware-accelerated Base64 decoding routines on x86-64 processors supporting AVX-512 instructions. Calif demonstrated unauthenticated remote code execution against default installations of Jenkins, Atlassian Jira, Confluence, Keycloak, PingFederate, and JFrog Artifactory, as well as authenticated attacks on tools like Apache Kafka Broker and Elasticsearch. There is no universal exploit because attacks require application-specific payload delivery. Until patches are applied, administrators can mitigate the issue by disabling hardware-accelerated Base64 or turning off AVX-512 in the Java runtime. [1][3][4]
Calif researcher Quang Luong identified the bug using OpenAI's Codex coding tool in roughly an hour, spending under $200 in computing tokens. The United States Cybersecurity and Infrastructure Security Agency urged organizations to deploy fixes immediately. While Cisco Talos reported no observed in-the-wild exploitation before disclosure, it cautioned that scanning and attack activity are expected to rise quickly. [1][3][4]