ServiceNow CVE-2026-6875 Exploited: Patch AI Platform RCE

ServiceNow CVE-2026-6875 has moved from a critical patch advisory into a practical incident-response priority after threat intelligence researchers reported in-the-wild exploitation attempts against the ServiceNow AI Platform. ServiceNow says it is not currently aware of exploitation against ServiceNow instances in its own advisory, and later told security press that it had not observed evidence tied to ServiceNow-hosted instances, but the safest path for customers is still immediate patch verification and log review.^[1][2]

The vulnerability is a sandbox escape and code-injection flaw in the ServiceNow AI Platform. The CVE record says an unauthenticated user, in certain circumstances, could execute code within the platform, and ServiceNow assigned the issue a CVSS v4.0 score of 9.5 Critical.^[1] That combination matters because ServiceNow instances often sit close to IT service management, workflow automation, identity-adjacent processes, and integrations that can reach internal systems.

Searchlight Cyber, which reported the bug to ServiceNow on April 1, published the technical root cause after coordinated remediation. Its write-up describes an unauthenticated path that starts from user-controlled query input, escapes ServiceNow’s stricter script sandbox, and can lead to broad instance compromise and command execution through connected proxy servers in some environments.^[3] The useful defender takeaway is not the exploit chain; it is that attempts to block only one public proof-of-concept route are weaker than applying the platform fixes.

ServiceNow says hosted instances received mitigations quickly after the April report, and the CVE record says updates were also provided to self-hosted customers and partners.^[1] The affected release scope listed for CVE-2026-6875 includes versions before Australia Patch 2, Yokohama Patch 12 Hot Fix 1b or Yokohama Patch 13, Zurich Patch 7b or Zurich Patch 9, and the Brazil EA/GA fixed line.^[1]

What ServiceNow admins should do now

First, confirm whether every ServiceNow environment is on a fixed family release or has the relevant vendor security update applied. Do not assume the production hosted tenant is the only exposure: self-hosted, partner-managed, test, cloned, integration, and internet-reachable non-production instances are the usual places where patch drift survives.

Second, review web access and application logs for unusual requests to /assessment_thanks.do, suspicious pre-authentication activity, abnormal script evaluation errors, unexpected admin-user creation, and changes around integrations or proxy/MID-style connectors. Help Net Security and BleepingComputer both report that Defused observed attempts hitting the same pre-authentication endpoint described by Searchlight Cyber, while using a different route to reach the code-execution primitive.^[2][4] That is exactly why allowlisting, WAF signatures, or one-off request blocking should be treated as temporary visibility aids, not remediation.

Third, treat suspicious activity as a credential and integration review, not only a web-server alert. If an instance was exposed before patching, check recent privileged account changes, OAuth/app credentials, outbound integrations, workflow modifications, and any connected systems reachable from ServiceNow. The same pattern has shown up across other AI/workflow incidents HowToFix has covered, including exposed Langflow AI workflow servers, the Hugging Face autonomous AI breach, and NadMesh attacks against AI services and cloud keys.

For most organizations, the priority order is straightforward: verify the fixed build, identify any lagging self-hosted or partner-managed instances, preserve logs before retention rolls them off, and rotate credentials tied to exposed integrations if there is any sign of abnormal activity. The public evidence currently supports a cautious wording: exploitation activity has been reported by external researchers, while ServiceNow has not confirmed compromise of hosted instances. That distinction should not slow patching.

References

1. NIST NVD. “CVE-2026-6875 Detail.” Published July 13, 2026; last modified July 14, 2026. https://nvd.nist.gov/vuln/detail/CVE-2026-6875
2. Help Net Security. “ServiceNow pre-auth RCE exploited in the wild (CVE-2026-6875).” July 20, 2026. https://www.helpnetsecurity.com/2026/07/20/servicenow-cve-2026-6875-exploited/
3. Searchlight Cyber. “Smashing the ServiceNow Sandbox: Pre Authentication RCE.” July 14, 2026. https://slcyber.io/research-center/smashing-the-servicenow-sandbox-pre-authentication-rce/
4. BleepingComputer. “Critical ServiceNow code execution flaw now exploited in attacks.” July 20, 2026. https://www.bleepingcomputer.com/news/security/critical-servicenow-code-execution-flaw-now-exploited-in-attacks/