Skip to main content
Mallory
CriticalPublic exploitVulnerability

CVE-2026-6875: Pre-authentication sandbox escape RCE in ServiceNow AI Platform

Published Jul 13, 2026 · Updated 2 days ago
Trending this weekNot trendingOutside this week’s top 100
Mention activity · 30 days1 mention · peak 1 in a dayToday
CVSS9.5Critical · v4
EPSS78.5%100th percentile
CISA KEVNoNot listed
In the wildReported15 exploitation records in Mallory
IdentifiersCVE-2026-6875CWE-94 · Improper Control of Generation of…

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.

Do you run this?See affected versions in your environment and whether adversaries are exploiting this vulnerability against assets like yours.
Check your exposure
Analyst brief

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.

Public exploits

Exploits

2 exploits validated by Mallory's analysis.

View more in app →
CVE-2026-6875Small standalone PoC repository with 4 files: license, gitignore, README, and a single Python exploit script. The only code file, cve-2026-6875.py, is the clear entry point and implements a web/network exploit against a claimed vulnerable ServiceNow endpoint. The script accepts a target URL and optional collaborator URL, builds a malicious javascript: payload, URL-encodes it, and submits it via HTTP POST in the sysparm_assessable_type parameter. The payload is designed to abuse ServiceNow server-side objects and methods such as DiscoveryFunctions, AbstractAjaxProcessor, Class.create, and GlideHTTPRequest to escape a sandbox and execute arbitrary JavaScript in global context. If a collaborator URL is supplied, the payload causes the target to make an outbound GET request with a distinctive X-Exploit header, providing an observable success signal; otherwise it executes a simple gs.print marker. The script disables TLS verification, uses a browser-like User-Agent, and prints response metadata for operator feedback. README content is mostly generic legal/educational boilerplate and does not add operational exploit logic. Overall, this is a basic but functional standalone PoC rather than a framework module or detection script.HORKimhab · Disclosed Jul 21, 2026 · python · markdown · web · networkOperationalCVE-2026-6875-PoC-ExploitThis repository is a Python-based exploit toolkit for the claimed ServiceNow pre-auth RCE vulnerability CVE-2026-6875. It contains two main code files: exploit.py, which is the offensive component, and verifier.py, which is a safer vulnerability scanner/verifier. The repository is not tied to a major public framework like Metasploit or Nuclei; it is a standalone tool. exploit.py is the primary entry point and implements a full attack chain against ServiceNow over HTTPS. The core logic builds a URL-encoded javascript: payload for the sysparm_assessable_type parameter on /assessment_thanks.do. The payload attempts a sandbox escape by abusing DiscoveryFunctions, AbstractAjaxProcessor, and Class.create, then executes OS commands through java.lang.Runtime.getRuntime().exec(). The code includes payload builders for direct command execution, callback-based output exfiltration using ServiceNow HTTP request classes, and reverse-shell generation. It also exposes many operator-focused CLI options such as single-target mode, file-based batch mode, threading, retries, timeout/delay tuning, custom cookie and User-Agent values, insecure TLS, WAF bypass obfuscation, callback host/port, webhook, persistence, lateral movement, cleanup, fileless mode, and C2 host/port. Based on these features, the exploit is more than a simple PoC and is best classified as OPERATIONAL. verifier.py is a separate detection-oriented utility. It fingerprints ServiceNow by requesting /navpage.do, /login.do, /index.do, and /ssologin.do; attempts rough version inference from response content; checks whether /assessment_thanks.do is reachable; and tests for JavaScript injection using benign gs.print-style payloads. It appears intended to validate exposure without full weaponized exploitation. Fingerprintable targets and observables are centered on HTTPS ServiceNow endpoints, especially /assessment_thanks.do and the sysparm_assessable_type parameter. The exploit also supports attacker infrastructure endpoints for callbacks or reverse shells, represented by operator-supplied callback/C2 host and port values. Additional filesystem observables mentioned in the repository include /var/tmp/, /etc/passwd, /etc/shadow, ~/.ssh/authorized_keys, /etc/systemd/system/, and extraweb_access.log, which are relevant as post-exploitation artifacts or IOC locations. Overall, the repository's purpose is to provide both a vulnerability verifier and a weaponized exploitation tool for unauthenticated remote code execution against vulnerable self-hosted ServiceNow instances.tc4dy · Disclosed Jul 21, 2026 · python · web · networkOperational
FAQ

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.
Activity feed

Recent activity

175 sources tracked across advisories, community write-ups, and news. New activity surfaces here as Mallory finds it.

View more in app →
+22 more in the timeline. Sign up to see them
Do you run this?See affected versions in your environment and whether adversaries are exploiting this vulnerability against assets like yours.
Check your exposure
Mallory platform

The version that knows your environment.

This page is what’s public. Mallory adds the parts that aren’t: which of your assets are affected, which adversaries are exploiting it right now, and which detections to deploy tonight.
Exposure mappingQuery your assets running an affected version, and investigate the blast radius.
Threat actor evidenceEvery report linking this CVE to a named adversary.
Associated malwareMalware families linked to this CVE, with evidence and IOCs.
Detection signatures1YARA, Sigma, Snort, and vendor rules, auto-deployed to your SIEM.
Vendor-by-vendor mappingCross-references every affected SKU, including bundled OEM variants.
Social activity148Community discussion across Reddit, Mastodon, and other social sources.