An urgent security threat has emerged for organizations utilizing the ServiceNow AI Platform. On July 20, 2026, threat intelligence firm Defused confirmed that attackers have begun actively exploiting CVE-2026-6875 in the wild. This critical vulnerability allows unauthenticated, remote attackers to escape the platform's script sandbox and execute arbitrary code on vulnerable instances. Organizations running self-hosted or on-premises deployments are urged to patch immediately, as threat actors have already adapted public proof-of-concept (PoC) code to bypass standard mitigations.

Chronology of the Threat

The lifecycle of CVE-2026-6875 began on April 1, 2026, when security researchers at Searchlight Cyber responsibly reported a critical sandbox-escape vulnerability to ServiceNow. Because fixing a sandbox escape in an AI and automation platform without breaking core workflows is complex, ServiceNow spent nearly three months engineering a clean resolution.

While ServiceNow successfully patched its hosted cloud environments within 24 hours of discovery, compiling and testing the patch for self-hosted instances took until July 13, 2026, when ServiceNow officially released the security updates alongside a detailed public advisory.

Unfortunately, the window of safety closed rapidly. Following the release of a technical write-up and a proof-of-concept (PoC) exploit, threat intelligence monitors detected the first real-world exploitation attempts on Friday, July 17, 2026. By July 20, 2026, active exploitation in the wild was confirmed. Notably, threat actors have already modified their exploit payloads, utilizing a different execution route than the published PoC to bypass five distinct mitigations that ServiceNow had previously placed in its code.

Technical Impact and Severity

With a near-maximum CVSS score of 9.5, CVE-2026-6875 is classified as a critical-severity bug. The flaw is a code injection vulnerability residing in the ServiceNow AI Platform (formerly the Now Platform). Specifically, it affects the Brazil, Australia, Zurich, and Yokohama releases.

Because it is a pre-authentication vulnerability, attackers do not need to possess valid user credentials or undergo any authentication to exploit it. Attackers target a specific pre-authentication endpoint—`/assessment_thanks.do`—and input crafted payloads designed to break out of the platform’s Javascript script sandbox.

Once the sandbox is bypassed, the attacker gains the ability to execute remote commands on the underlying host. The consequences of a successful compromise are severe. An attacker can gain full control over the ServiceNow instance, access highly sensitive corporate databases, manipulate automated digital workflows, and compromise connected proxy servers or internal directory systems.

Remediation and Defense Strategies

Because attackers are actively tweaking their exploitation techniques to bypass standard network signatures, immediate patching is the only complete defense. IT administrators should implement the following steps:

1. Apply the July 13 Patch: Self-hosted and on-premises administrators must immediately update their ServiceNow instances to the patched versions of the Brazil, Australia, Zurich, or Yokohama releases. 2. Enable Guarded Script: The latest security updates introduce "Guarded Script," a feature designed to strictly restrict the types of code permitted to execute within sandbox environments. Ensure this feature is active. 3. Analyze Endpoint Logs: Inspect web server logs and web application firewall (WAF) logs for anomalous POST or GET requests targeting the `/assessment_thanks.do` URI, especially those originating from unrecognized external IP addresses. 4. Implement Zero-Trust Access: Restrict public internet exposure of your ServiceNow management portal. Hide the interface behind a Virtual Private Network (VPN) or a Zero-Trust Network Access (ZTNA) gateway to prevent unauthenticated external scanning.

---