CVE-2026-12057: Foxit Ai RCE Vulnerability Explained

CVE-2026-12057 Overview

CVE-2026-12057 affects Foxit AI, where the application fails to fully intercept dangerous interfaces when executing JavaScript embedded in PDF files within its sandbox. Attackers can craft a malicious PDF that loads remote scripts, bypassing sandbox restrictions and achieving arbitrary code execution on the local system. The flaw is categorized under [CWE-829: Inclusion of Functionality from Untrusted Control Sphere]. Exploitation requires a user to open the malicious PDF, but no privileges are needed. Successful exploitation grants full compromise of confidentiality, integrity, and availability on the affected host.

Critical Impact
A user opening a crafted PDF in Foxit AI can trigger remote script loading, leading to arbitrary code execution under the user's context.

Affected Products

Foxit AI (all versions prior to the vendor-released fix)
Vendor: Foxit
Component: foxit:ai

Discovery Timeline

2026-06-15 - CVE-2026-12057 published to NVD
2026-06-17 - Last updated in NVD database

Technical Details for CVE-2026-12057

Vulnerability Analysis

Foxit AI embeds a JavaScript engine to support scripted PDF features. The engine runs inside a sandbox that is intended to block access to sensitive interfaces such as filesystem, network, and external resource loaders. The sandbox fails to intercept a subset of these dangerous interfaces, allowing scripts inside a PDF to reach out and load additional code from remote locations. Once remote script content is loaded into the privileged context, the attacker can execute arbitrary code with the rights of the user running Foxit AI.

The vulnerability falls under [CWE-829], which describes the inclusion of functionality from an untrusted control sphere. Because the loaded script originates from an attacker-controlled remote endpoint, the executing application effectively runs adversary-supplied logic. EPSS data indicates a current exploitation probability of 0.126%, and no public proof-of-concept has been published at this time.

Root Cause

The root cause is an incomplete sandbox policy. Specific JavaScript interfaces that can initiate outbound network requests or dynamically load script content were not added to the deny list. The sandbox enforces restrictions on some APIs but leaves others reachable, breaking the security boundary that the sandbox is designed to provide.

Attack Vector

The attack vector is local and requires user interaction. An attacker crafts a PDF containing embedded JavaScript that calls one of the unrestricted interfaces. The PDF is delivered through email, file sharing, or a website. When the victim opens the PDF in Foxit AI, the embedded script executes, fetches an attacker-hosted payload, and runs it. The attacker gains code execution as the local user, enabling credential theft, persistence, or lateral movement.

No verified exploit code is publicly available. See the Foxit Security Bulletins for vendor-supplied technical details.

Detection Methods for CVE-2026-12057

Indicators of Compromise

Foxit AI processes spawning child processes such as cmd.exe, powershell.exe, or wscript.exe shortly after opening a PDF.
Outbound HTTP or HTTPS connections originating from Foxit AI to uncommon or recently registered domains.
PDF files containing /JS or /JavaScript objects that reference external URLs or use dynamic script-loading functions.
New executable files or scripts written to user-writable directories immediately following a PDF open event.

Detection Strategies

Hunt for parent-child process relationships where a Foxit AI process is the parent of a scripting host or shell.
Inspect PDF samples with static parsers to flag embedded JavaScript that calls network or eval-style interfaces.
Correlate PDF open events with subsequent unusual outbound network traffic from the same user session.

Monitoring Recommendations

Enable endpoint process and command-line logging on systems where Foxit AI is installed.
Forward DNS and proxy logs to a centralized analytics platform to identify suspicious destinations contacted by Foxit AI.
Alert on any write of executable content to %APPDATA%, %TEMP%, or user profile directories by Foxit AI.

How to Mitigate CVE-2026-12057

Immediate Actions Required

Apply the vendor-released update for Foxit AI as documented in the Foxit Security Bulletins.
Block opening of PDF attachments from untrusted senders at the email gateway until patching is complete.
Restrict outbound network access from Foxit AI processes where business workflows allow.
Educate users to avoid opening PDFs delivered by unsolicited messages or untrusted links.

Patch Information

Foxit has acknowledged the issue through its security bulletin portal. Administrators should consult the Foxit Security Bulletins page for the fixed version of Foxit AI and apply it across all installations. Verify that the deployed build matches or exceeds the fixed version listed in the advisory.

Workarounds

Disable JavaScript execution within Foxit AI through the application's preferences until the patch is installed.
Use application allow-listing or endpoint controls to prevent Foxit AI from spawning shell or scripting host processes.
Apply network egress filtering so Foxit AI cannot retrieve content from arbitrary internet hosts.

bash

# Configuration example: disable JavaScript in Foxit AI via registry (Windows)
reg add "HKCU\Software\Foxit Software\Foxit AI\Preferences\JavaScript" /v bEnableJS /t REG_DWORD /d 0 /f