A critical ServiceNow AI Platform vulnerability tracked as CVE-2026-6875 is reportedly being exploited after technical details and proof-of-concept code became public. Threat intelligence company Defused Cyber reported observing exploitation in the wild, while ServiceNow told SecurityWeek that its investigation had found no evidence connecting the reported activity to instances hosted by the company. Self-hosted customers remain a priority for immediate patching.

ServiceNow disclosed the flaw on July 13 and assigned it a critical CVSS 4.0 base score of 9.5. The company said an unauthenticated user could, under certain circumstances, execute code within the ServiceNow platform. ServiceNow deployed a security update to hosted instances and supplied fixes to self-hosted customers and partners. Its original advisory said the company was not aware of exploitation at that time.

The vulnerability is a sandbox escape associated with CWE-94, improper control of code generation. Searchlight Cyber, which reported the issue to ServiceNow on April 1, said successful exploitation could compromise a ServiceNow instance and connected proxy servers. Public reporting describes attack traffic reaching the pre-authentication endpoint /assessment_thanks.do through HTTP POST requests and using the sandbox-escape technique to reach arbitrary code execution.

Defused initially reported that the observed payload used a different path to the same execution primitive described by Searchlight Cyber. The firm later corrected that assessment, saying closer review showed the captured payload matched the published proof of concept. SecurityWeek noted that no separate exploitation reports were available and that the observed activity could involve security researchers scanning or testing exposed systems, leaving the identity and intent of the operators unconfirmed.

Affected releases include versions before Australia Patch 2; Yokohama Patch 12 Hot Fix 1b and Patch 13; Zurich Patch 7b and Patch 9; and Brazil EA and GA. The National Vulnerability Database lists a network attack vector, no required privileges or user interaction, high attack complexity, and potentially severe effects on confidentiality, integrity and availability. NVD is still awaiting its own enrichment assessment.

Administrators should confirm that their instance is on a fixed family release or later, prioritize self-hosted deployments, and review web and application logs for unusual POST requests to the identified endpoint. ServiceNow customers should follow vendor guidance and contact support if patching assistance is required. Defenders should treat public exploit availability as increased exposure while avoiding assumptions about confirmed criminal campaigns until additional evidence emerges.