CVE-2026-109331: Network-AI security control bypass in ClaudeHookBridge

CVE-2026-109331 · Severity: high · CVSS 7.4 · Published 2026-10-12

Network-AI is a library that integrates Claude AI agents with code execution capabilities. A security control bypass in the ClaudeHookBridge component allows attackers to evade operator-configured deny patterns by placing dangerous payloads (web shells, remote code execution scripts) in non-primary fields, bypassing the intended security gate. An attacker with prompt injection access can exploit this to execute arbitrary code despite configured restrictions.

Technologies: Jovancoding Network-AI. Vendors: Jovancoding.

Quick answers

Is CVE-2026-109331 exploited in the wild?
No. Junglewise Threat Intelligence has no report of CVE-2026-109331 being exploited in the wild.
How severe is CVE-2026-109331?
High severity, with a CVSS score of 7.4, published 12 October 2026.
Which products and versions are affected?
Jovancoding Network-AI 5.14.0 through 5.15.1.
Is there a fix for CVE-2026-109331?
Yes: fixed in 5.15.2. Junglewise finds every place you run it, drafts the upgrade and applies it once you approve.
How is CVE-2026-109331 attacked?
It is attacked with local access, in a software library.
Am I affected by CVE-2026-109331?
Check whether you run Jovancoding Network-AI at an affected version. Junglewise does that check for you around the clock: connect your code, cloud and domains, and it matches new vulnerabilities like this one against what you actually run and tells you when one reaches you.

How to fix

  • Fixed in: 5.15.2

Technical details

The extractFullTarget function inspects only the first present candidate field (command, file_path, filePath, path, url, query, pattern, prompt) from tool input when matching against deny patterns, rather than examining all fields or the full tool input JSON. An attacker can place executable content in secondary fields (e.g., Write.content, Edit.new_string, mcp script parameters) while keeping the first candidate field benign, causing the deny pattern match to fail against the wrong target string. The vulnerability affects versions 5.14.0 through 5.15.1, with a fix in 5.15.2 that matches deny patterns against all candidate fields.

Affected products

  • Jovancoding Network-AI 5.14.0 through 5.15.1

Timeline

  • 2026-09-27: disclosed
  • 2026-10-12: advisory

References

This page as Markdown. Free to reuse under CC BY 4.0 with a link back.