Vulnerability in the ServiceNOW AI Platform

Overview

A critical vulnerability in the ServiceNow AI platform, tracked as CVE‑2026‑6875, has been publicly disclosed with a working proof‑of‑concept (PoC). The flaw enables pre‑authentication remote code execution (RCE), allowing attackers to escape the restricted script sandbox and run arbitrary code on vulnerable ServiceNow instances.

How the Exploit Works

Researchers from Assetnote at Searchlight Cyber detailed the attack in their report “Smashing the ServiceNow Sandbox: Pre‑Authentication RCE.”

• The vulnerability lies in the GlideRecord query API, which processes user‑controlled input.
• Malicious JavaScript values can be evaluated before query execution.
• Although ServiceNow uses a restricted sandbox to block dangerous functions (like eval or dynamic function creation), attackers bypassed these restrictions via the gs.include() function.
• By manipulating global JavaScript objects and constructor behavior, attackers forced code execution in a less restricted context, escaping the sandbox.

Once outside the sandbox, attackers gain access to internal APIs, enabling:

• Database access.
• Administrator account creation.
• Command execution on MID Servers, which connect ServiceNow cloud instances to enterprise networks.

Impact & Risks

Successful exploitation could lead to:

• Full compromise of ServiceNow instances.
• Theft or manipulation of sensitive data stored in tables.
• Unauthorized administrative actions.
• Broader enterprise compromise via MID Server proxies.

Although ServiceNow stated it is not aware of active exploitation, the release of a public PoC significantly raises the risk of opportunistic attacks.

ServiceNow’s Response

• Immediate cloud‑side mitigations were deployed within 24 hours of disclosure.
• Broader fixes introduced Guarded Script protections, reducing the attack surface of sandboxed code.
• Under this model:
  • Simple expressions and basic function calls remain supported.
  • Complex scripts (variables, loops, conditionals) must be moved into Script Includes.

Patched Versions:

• Brazil EA & GA
• Australia Patch 2
• Zurich Patch 7b & Patch 9
• Yokohama Patch 12 Hot Fix 1b & Patch 13

ServiceNow customers should urgently confirm they are running a patched release and review the Incompatible Guarded Scripts list after upgrading.

Defensive Recommendations

Organizations should:

• Patch ServiceNow instances immediately to the latest supported release.
• Audit MID Server configurations for potential exposure.
• Review custom scripts to ensure compatibility with Guarded Script protections.
• Monitor for suspicious activity, especially unusual API calls or privilege escalations.

Expert in the Cloud Insight

This disclosure highlights the fragility of sandbox mechanisms in enterprise platforms. Attackers increasingly exploit overlooked script‑execution paths to bypass restrictions. For defenders, the lesson is clear: sandbox escapes are high‑impact vulnerabilities, requiring immediate patching, strict monitoring, and governance of custom scripts.