Java4Shell Fixes Arrive After AI-Assisted Discovery

Java4Shell Fixes Arrive After AI-Assisted Discovery

OpenJDK and Oracle released fixes on Nov. 17 for a security flaw called Java4Shell. The bug can let an attacker run code in an affected Java application without logging in. Security firm Calif said it used AI to find the vulnerability and build its first working exploit in about three hours.

Key facts about the flaw

  • The bug is tracked as CVE-2026-114938. It affects unpatched OpenJDK versions starting with 17.0.6.
  • Only x86-64 processors that support AVX-512 VBMI instructions are affected. ARM64 systems are not affected.
  • Exposure also depends on whether an application passes attacker-controlled data to the Java Base64 decoder.
  • Red Hat said authentication mechanisms can pass Base64-encoded credentials and tokens to the decoder before a user logs in.
  • Calif and Wiz rate the flaw 9.2 out of 10 under CVSS 4.0. Oracle's CVE record uses the older CVSS 3.1 scale, rates it 8.1, and describes it as difficult to exploit.

How the bug works

Base64 is a way of representing binary data as text. OpenJDK speeds up decoding on supported processors. But the optimized code mishandles malformed input and writes beyond an allocated memory buffer, according to Calif's technical account. Java can reject the bad input only after nearby memory has already been corrupted.

The risk reaches beyond any single application because the flaw sits in the Java runtime itself, the software layer many programs rely on.

What security firms say AI did

Wiz said its Atlas AI vulnerability-research system built a proof-of-concept exploit within two hours. In a separate statement, the company said its team confirmed remote code execution against Jira, Confluence and Jenkins. Wiz cautioned that adapting an exploit to an arbitrary target is difficult.

Calif described a researcher-guided process. Quang Luong directed Codex to audit selected parts of OpenJDK, discarded early findings that were unreachable or unrealistic, and eventually received the Base64 finding. Calif said discovery took an hour and the first working exploit another two hours. Its researchers also built a heap tracer and visualizer to help AI agents when exploit development stalled.

OpenAI co-founder Greg Brockman shared a New York Times report on the finding and credited GPT-6 Cyber. He said the work was part of OpenAI's Patch the Planet initiative.

Calif emphasized there is no universal exploit. Each application needs a tailored approach. The firm said it is withholding application-specific exploit chains until users have had a chance to upgrade, and will publish them later.

Which versions are affected

OpenJDK's advisory lists affected release families 17, 21, 25 and 27. Oracle also lists its Java SE 8u511b33-perf build as affected. That does not mean every Java 8 distribution is vulnerable.

What operators should do

Oracle's Nov. 17 security releases include JDK 17.0.21.1+4, 21.0.13.1+4, 25.0.5.1+4 and 27.0.1.1+4. Operators should use their own distribution's corresponding security update rather than assume Oracle's build numbers apply to every vendor. Red Hat said it was expediting fixes and urged customers to update as errata become available.

For deployments still waiting on an update, Red Hat gives two interim startup options:

  • -XX:-UseBASE64Intrinsics disables hardware-accelerated Base64 processing. Red Hat prefers this narrower option because it keeps other AVX-512 optimizations enabled.
  • -XX:UseAVX=2 disables AVX-512 more broadly. Red Hat warns it should only be used on x86-64, because ARM64 JVMs do not recognize it and will fail to start.

Wiz said the workarounds may affect performance. It recommended restricting public access to vulnerable management portals until updates or mitigations are in place. For products that bundle their own JVM, Red Hat recommends following the product vendor's supported update path instead of replacing the runtime independently.

Sources

Newisty Editorial Team
Written by

Newisty Editorial Team

Technology · Crypto · Digital Economy
View all posts

Newisty Editorial Team covers technology, cryptocurrency, digital products, online platforms, developer tools and the wider digital economy. Our content is researched from official sources, company announcements, public documentation, market data and other primary or reputable sources. Articles are reviewed and edited before publication for clarity, accuracy and useful context.

Comments (0)

Leave a comment
Your comment will appear publicly after submission.
No comments yet. Be the first to comment!