Active Exploitation: Attackers Target Critical ServiceNow AI Platform Code Execution Flaw
Executive Summary
On July 20, 2026, threat intelligence firm Defused confirmed that malicious threat actors have begun actively exploiting a critical-severity vulnerability in the ServiceNow AI Platform in real-world attacks. Tracked as CVE-2026-6875, this vulnerability is a high-severity sandbox escape and unauthenticated remote code execution (RCE) flaw affecting ServiceNow enterprise instances running AI workflow and automation modules. The flaw allows remote, unauthenticated attackers to escape the restricted scripting execution environment and execute arbitrary OS commands directly on backend application infrastructure, exposing critical enterprise databases, customer records, and internal IT service workflows to full compromise.
Deep-Dive Technical Analysis
The attack vector leverages an architectural flaw within the platform's custom JavaScript execution engine utilized for AI agent orchestration and task execution:
1. Unauthenticated API Endpoint Exposure: The platform exposes specific REST API endpoints responsible for dynamic script compilation and AI intent evaluation. Certain legacy endpoints failed to enforce mandatory OAuth authentication tokens or access control checks, allowing external network callers to submit raw payload strings.
2. Sandbox Escape via Reflection and Prototype Pollution: The ServiceNow AI platform uses an isolated JavaScript sandbox to evaluate user scripts safely. However, CVE-2026-6875 allows attackers to bypass variable sanitization by exploiting object prototype pollution and Java reflection classes accessible within the scripting runtime. By invoking Java system methods through contaminated prototypes, the script breaks out of the JS sandbox.
3. Arbitrary Command Execution: Once outside the sandbox context, the attacker gains direct access to underlying OS process builders (e.g., java.lang.Runtime.getRuntime().exec()), enabling the execution of shell scripts, reverse shells, and database exfiltration commands with host-level application privileges.
Industry Impact and Recommendations / Mitigations
Given the central role ServiceNow plays in enterprise IT governance, ticketing, and access management, active exploitation poses immediate systemic risk across Fortune 500 networks:
* Apply Vendor Security Patches Immediately: Organizations hosting or managing ServiceNow instances must immediately apply hotfixes and family patches released by ServiceNow targeting CVE-2026-6875.
* Implement Web Application Firewall (WAF) Rules: Deploy custom WAF detection signatures to monitor incoming API traffic for prototype pollution keywords, Java reflection signatures, and OS command syntax targeting ServiceNow endpoints.
* Restrict Access & Restrict Service Accounts: Ensure that ServiceNow backend server instances are restricted behind zero-trust network access (ZTNA) gateways and that the application daemon runs under a strictly restricted, non-root Linux user service account.
References:
* BleepingComputer: Critical ServiceNow code execution flaw now exploited in attacks
* SecurityWeek: ServiceNow AI Platform RCE Vulnerability