A Guth Labs publication ↗
Back to News

Calif discloses AI-assisted discovery of Java4Shell flaw

Green glass shield protecting a cream cup with a gold repaired seam.
AI-generated editorial illustration

Security firm Calif disclosed Java4Shell on November 17, saying AI-assisted research uncovered a memory-corruption flaw in OpenJDK that can enable remote code execution in affected applications.

In its technical account, researcher Quang Luong describes directing Codex toward parts of the Java runtime and filtering out unusable findings before identifying the bug. Calif says discovery took an hour and the first working exploit another two hours, with token costs below $200. Those figures are the research firm’s account, not independently measured productivity results.

The vulnerability, CVE-2026-114938, affects an optimized Base64 decoding path. Exposure depends on the Java build, application and hardware: Calif says the vulnerable code requires x86-64 processors with AVX-512 VBMI support, and ARM64 processors are not affected.

Oracle’s November security advisory lists the flaw, credits Luong and Albert Yang, and rates it 8.1 under CVSS 3.1. Calif says Oracle released a patch on November 17 and urges users to update.

The case illustrates AI’s role in expert-led security research, not autonomous discovery without human judgment. Calif is withholding application-specific exploit chains while users upgrade.

Sources and citations

Revision history

Published .