ServiceNow Patches Critical AI Platform Flaws Enabling Data Theft

ServiceNow released patches on 24 September 2026 for five vulnerabilities in its AI Platform, affecting the Yokohama, Zurich and Australia release branches before the September patch cycle. The issues comprise two critical and three high-severity flaws. The most serious are CVE-2026-13016, an unauthenticated SQL injection vulnerability rated 9.3 under CVSSv4, and CVE-2026-86860, an unauthenticated sensitive-data disclosure flaw also rated 9.3.

The remaining vulnerabilities are CVE-2026-86858, an unauthenticated privilege-escalation issue via GraphQL (8.7); CVE-2026-86859, an unauthenticated arbitrary-record disclosure flaw (8.7); and CVE-2026-86857, an authorisation bypass (8.4).

According to ServiceNow, the vulnerabilities were found through internal assessments and bug-bounty programmes, and the company has not identified evidence of malicious exploitation. The article also says no public proof-of-concept code has been published.

Nevertheless, the flaws could allow unauthenticated attackers to submit malicious database queries, extract instance data or gain additional privileges, while CVE-2026-86857 could let authenticated users access restricted datasets through inadequate authorisation checks.

ServiceNow has reportedly updated cloud-hosted instances automatically. Administrators of self-hosted deployments should apply the security update for their release branch. The listed fixes include Yokohama Patch 13 Hot Fix 5a and Zurich Patch 10 Hot Fix 3b; the table also lists Zurich Patch 10 Hot Fix 4a W32 (+4). The article advises applying the latest available update for affected Australia deployments as well.