CVE-2026-6875: Pre-authentication sandbox escape RCE in ServiceNow AI Platform
CVE-2026-6875 is a code-injection and sandbox-escape vulnerability in the ServiceNow AI Platform that permits unauthenticated remote code execution under certain circumstances. Pre-authentication application logic passes attacker-controlled input to GlideRecord.addQuery(), where JavaScript-prefixed query values are evaluated inside a restrictive filter-script sandbox. The gs.include() mechanism evaluates included scripts in a context without the same restrictions. By manipulating global JavaScript properties consumed during script inclusion, an attacker can cause attacker-controlled code to be compiled and executed outside the restrictive sandbox. Successful exploitation can compromise instance data and administrative functionality and potentially extend to configured MID or proxy servers.
Impact, mitigation & remediation
What it means and what to do now: patch path, interim mitigations, and the assume-compromise checklist.
Impact
What an attacker gets, and what they’ve been doing with it.
Successful exploitation can fully compromise a ServiceNow instance, permitting arbitrary table-data access, unauthorized data modification, and creation of administrator accounts. Configured MID or proxy servers may also be used to execute shell commands, potentially extending compromise into internal enterprise networks. Initial claims of malicious in-the-wild exploitation were subsequently qualified after captured activity was identified as matching a public proof of concept; ServiceNow stated it had not observed evidence connecting that activity to its hosted instances.
Mitigation
If you can’t patch tonight, do this now.
Enable and verify Guarded Script enforcement to restrict sandboxed input to a single simple expression and prohibit assignments, declarations, control flow, and multiple statements. Self-hosted deployments require administrators to advance enforcement phases manually. If patching is delayed, restrict access to exposed pre-authentication interfaces and segment connected MID or proxy infrastructure. Monitor for anomalous unauthenticated requests, unexpected administrator creation, unusual table access, and MID Server command execution. These measures do not replace patching.
Remediation
Patch, then assume compromise.
Apply ServiceNow's security updates for CVE-2026-6875 or upgrade to an applicable patched release, verifying the complete release, patch, hot-fix, and branch level against advisory KB3137947. Listed fixes include Australia Patch 2, Yokohama Patch 12 Hot Fix 1b or Patch 13, and Zurich Patch 7b or Patch 9; fixes were also supplied for Brazil EA and GA. ServiceNow deployed protections to hosted instances, but customers should confirm update status. Self-hosted customers and partners must apply the supplied updates themselves.
Exploits
2 exploits validated by Mallory's analysis.
Frequently asked questions
- Is CVE-2026-6875 exploited in the wild?
- Yes. As of Nov 15, 2026, Mallory has reports of exploitation in the wild.
- Is there a public exploit for CVE-2026-6875?
- Yes. Mallory tracks 2 public exploits or proofs of concept for CVE-2026-6875, the earliest disclosed Jul 21, 2026.
- Is there a patch for CVE-2026-6875?
- Apply ServiceNow's security updates for CVE-2026-6875 or upgrade to an applicable patched release, verifying the complete release, patch, hot-fix, and branch level against advisory KB3137947. Listed fixes include Australia Patch 2, Yokohama Patch 12 Hot Fix 1b or Patch 13, and Zurich Patch 7b or Patch 9; fixes were also supplied for Brazil EA and GA.
- How severe is CVE-2026-6875?
- CVE-2026-6875 has a CVSS v4 score of 9.5 (Critical) and an EPSS score of 78.5% (100th percentile) as of Nov 15, 2026.
Recent activity
175 sources tracked across advisories, community write-ups, and news. New activity surfaces here as Mallory finds it.