Oracle patches Java4Shell flaw after AI-assisted discovery
The vulnerability could allow attackers to compromise some Java applications, though it requires particular software and hardware. Researchers say Oracle’s security update fixes the flaw.

Review information is loading.
Security firm Calif has disclosed a vulnerability in OpenJDK, a widely used environment for running Java software. The firm says it used AI to help find the flaw and develop ways to exploit it.
Known as Java4Shell and tracked as CVE-2026-114938, the flaw lies in a part of OpenJDK that decodes Base64, a common way of representing data as text. The Canadian Centre for Cyber Security warned that successful exploitation could expose sensitive information, crash a process or allow an attacker to run code.
Not every Java application is vulnerable. Calif says the affected code requires a susceptible Java build and an x86-64 processor with a feature called AVX-512 VBMI. The application must also provide a route for specially crafted data to reach the decoder. ARM64 processors are not affected by the mechanism Calif described.
Calif says it demonstrated remote code execution in tests of several applications using their default configurations, including Jira and Jenkins without first logging in. The method differs between applications, it says. Another security firm, Upwind, reproduced memory corruption and a crash in its laboratory, but had not reproduced Calif’s reported code execution in those applications.
Oracle’s security advisory of 17 November lists the vulnerability and credits Calif researcher Quang Luong among those who reported it. Calif and Upwind say Oracle’s update fixes the flaw. The BBC reported on 18 November that Oracle had not acknowledged the bug, while also noting the security update.
Upwind identifies fixed Oracle builds for Java 17, 21, 25 and 27 as 17.0.21.1, 21.0.13.1, 25.0.5.1 and 27.0.1.1 respectively. Oracle also lists a particular Java 8 build and some GraalVM versions as affected. Organisations using other Java distributions should check their supplier’s updates. Calif and Upwind say disabling the accelerated decoder with `-XX:-UseBASE64Intrinsics` can provide a temporary mitigation.
Luong says he directed the AI coding assistant Codex to examine parts of OpenJDK and that it identified the memory flaw. Calif says AI also helped develop exploitation methods. In its alert on 17 November, the Canadian cyber security centre said it was not aware of active exploitation at that time.
Sign in to give this article a thumbs up or down.
From the discussion
Private test discussion. Comments are readers’ views and are not yet automatically fact-checked. Editing is available for 60 seconds after posting.
Sign in with a confirmed reader account and choose a username to read comments and participate.
Sorting applies to top-level comments; replies remain oldest first. New comments and likes can change the order. Refresh for the current ranking.
Loading comments…