ServiceNow patched four flaws, including three CVSS 10.0 vulnerabilities enabling unauthenticated remote code execution, and one sandbox escape rated 8.7.
On August 27, ServiceNow released a security advisory patching 4 vulnerabilities in the ServiceNow AI Platform and Now Platform, including 3 critical issues rated CVSS v4.0 10.0 and 1 sandbox escape vulnerability rated 8.7. The advisory states that under certain conditions, these issues could be remotely exploited by unauthenticated attackers, posing a direct threat to platform integrity and data security.
The CVEs patched this time include CVE-2026-18885, CVE-2026-18886, CVE-2026-74820, and CVE-2026-6876. On August 29, ServiceNow updated its guidance to say that no malicious exploitation of these vulnerabilities has been found so far, and it recommended that customers apply the updates as soon as possible or upgrade to the patched versions.
CVE-2026-18885 appears in the GraphQL Composite Data API and is a code injection vulnerability. According to the advisory, an attacker could execute arbitrary code under certain conditions and further access or modify instance data. From an attack-surface perspective, this type of issue usually means input data is not strictly validated and the backend processing logic may directly pass controllable content into a high-privilege execution flow, allowing the attacker to cross the original authorization boundary.
CVE-2026-18886 is related to improper access control in the system configuration image upload processor. The advisory says an attacker may be able to create or modify data beyond the originally permitted scope and obtain higher privileges. This shows that the core issue is not only the file upload itself, but an imbalance in the access checks around the upload process, turning a normally controlled operation into a privilege escalation path.
CVE-2026-74820 is an SQL injection vulnerability that allows an attacker to execute arbitrary SQL statements against the underlying database. If exploitation succeeds, the impact is not limited to data reading; it may also extend to data tampering, privilege-data poisoning, and even the stability of the overall platform.
CVE-2026-6876 is a sandbox escape vulnerability in the Now Platform, with a CVSS v4.0 score of 8.7. The advisory states that if successfully exploited, an attacker could execute arbitrary code on the Now Platform and gain platform access privileges beyond the originally permitted scope. Compared with a typical single-point flaw, the danger of a sandbox escape lies in the failure of isolation boundaries; once broken, the execution environment that should have been restricted can become a springboard for lateral expansion.
Overall, these 4 vulnerabilities show two high-risk themes: unauthenticated remote code execution and the breaking of database and privilege boundaries. If such flaws are combined with a predictable deployment environment, attackers may be able to achieve high-impact results with a low barrier to entry, especially since the downstream effects on an enterprise workflow platform often extend beyond a single system.
These vulnerabilities affect certain version lines of the ServiceNow AI Platform and Now Platform, including Xanadu, Yokohama, Zurich, and Australia. The advisory indicates that ServiceNow has deployed security updates for hosted instances and provided patches to partners and self-hosted customers.
In practice, affected parties include enterprises using hosted instances and self-hosted deployment environments that have not yet completed upgrades. Because the vulnerabilities involve code injection, SQL injection, and sandbox escape, the risk affects not only confidential data but may also impact platform access control, workflow execution, and system availability.
For defenders, the first step is to immediately confirm whether the environment falls within the affected version lines and compare it against the patched versions and update status provided by ServiceNow. If using hosted instances, verify that the vendor has completed the security update; if using a self-hosted environment, arrange upgrades and change validation as soon as possible.
Second, strengthen audits of API, upload components, and database query paths, especially modules involving GraphQL and the upload processor. If these components have insufficient input validation or weakened authorization checks, they can become high-risk attack entry points.
Third, it is recommended to continuously monitor for abnormal platform behavior, including unauthorized access attempts, suspicious SQL patterns, unusual data creation or modification activity, and signs of unexpected code execution. Even if no known malicious exploitation is currently observed, vulnerabilities with a perfect score should still be treated as a high priority.
Finally, after patching is complete, perform regression validation to confirm that core workflows, API access, and access controls have not introduced compatibility issues because of the update, avoiding new operational risks caused by the remediation itself.