First Seen: April 2026

Affected Products: ServiceNow AI Platform (formerly the Now Platform)

CVE-2026-6875 is a critical flaw in the ServiceNow AI Platform, a widely used service for building and automating business workflows. The cause is a code injection weakness: an attacker with no login can break out of the ServiceNow AI Platform's script sandbox and run their own code on a targeted instance. The impact is severe, because a successful attack can fully take over the affected ServiceNow AI Platform instance and any proxy servers connected to it, exposing sensitive data and disrupting critical workflows. A fix is available: ServiceNow patched its own hosted instances back in April 2026 and, on July 13, 2026, released updates for self-hosted customers and partners, so upgrading to the latest patched release closes the hole. Threat intelligence reports the flaw is now being exploited in the wild, while ServiceNow says it has not confirmed any exploitation of the instances it hosts.

CVE at a Glance

#1 — Sandbox-Escape Code Injection

CVE-2026-6875 is a critical flaw in the ServiceNow AI Platform that hands an unauthenticated attacker the ability to run their own code on a target instance. In plain terms, it is a code injection weakness, tracked as CWE-94 (Improper Control of Generation of Code): a crafted request reaches the ServiceNow AI Platform's server-side script engine and escapes the sandbox that is supposed to contain it, turning what should be harmless input into attacker-controlled code execution on the instance.

#2 — Reachable Without Authentication

The flaw is reachable before authentication. The exploit path targets a pre-authentication endpoint, /assessment_thanks.do, and from there abuses a sandbox-escape gadget to reach a code-execution primitive. This means an attacker who can send requests to an exposed ServiceNow AI Platform instance does not need valid credentials to attempt exploitation.

#3 — High Complexity, Severe Consequences

The vulnerability is described as exploitable only in high-complexity attacks, but with severe consequences when successful: unauthenticated code execution and full compromise of the ServiceNow AI Platform instance along with any proxy servers connected to it.

#4 — Affected Releases and Available Fixes

The vulnerability affects the ServiceNow AI Platform across multiple release families. ServiceNow addressed it in the following fixed releases: Australia Patch 2, Yokohama Patch 12 Hot Fix 1b, Yokohama Patch 13, Zurich Patch 7b, Zurich Patch 9, Brazil EA, and Brazil GA. Instances running versions before these fixed releases are affected. ServiceNow remediated its own hosted instances starting in April 2026 and made the corresponding security updates available to self-hosted customers and partners on July 13, 2026. The updates additionally ship a new feature called Guarded Script, which restricts the type of code that can run in sandbox contexts and is intended to make future sandbox escapes less likely.

#5 — Contested Exploitation Evidence

Evidence of active exploitation of CVE-2026-6875 is contested. A threat intelligence firm reported that the first exploitation attempts appeared on Friday, July 17, 2026, and confirmed active in-the-wild abuse over the following weekend. The observed payloads hit the same pre-authentication sink, but the attackers' sandbox-escape gadget reaches the same code-execution primitive through a different route than the published proof-of-concept. ServiceNow, by contrast, has not flagged the vulnerability as actively abused; its official advisory still states that it is not currently aware of exploitation against ServiceNow-hosted instances.

Affected Versions, CPE, and CWE

Patch Link

• https://support.servicenow.com/kb?id=kb_article_view&sysparm_article=KB3137947

References

• https://support.servicenow.com/kb?id=kb_article_view&sysparm_article=KB3137947
• https://x.com/DefusedCyber/status/2078418391321219448
• https://slcyber.io/research-center/smashing-the-servicenow-sandbox-pre-authentication-rce/