Attackers have begun exploiting CVE-2026-6875, a critical pre-authentication vulnerability in the ServiceNow AI Platform, according to threat intelligence firm Defused.
About the vulnerability
ServiceNow AI is a Platform-as-a-Service that lets organizations build and automate digital workflows.
CVE-2026-6875 is a code injection vulnerability that lets unauthenticated attackers escape ServiceNow’s script sandbox and execute code remotely on a targeted instance.
The vulnerability was unearthed by Searchlight Cyber researchers and reported to ServiceNow in early April 2026.
The latter company pushed out a security update to hosted instances the very next day, and made available patches and security updates to self-hosted customers and partners last week.
The existence of CVE-2026-6875 was publicly revealed on July 13. The security advisory and warning were followed by Searchlight Cyber’s very technical post detailing the flaw.
Researcher Adam Kues describes it as exploitable in high-complexity attacks, but allowing unauthenticated code execution and full compromise of the ServiceNow instance and any connected proxy servers.
Exploitation in the wild
Defused researchers say the first exploitation attempts appeared on Friday and confirmed active in-the-wild abuse over the weekend.
They said that the observed payloads hit the same pre-authentication endpoint (/assessment_thanks.do) that Searchlight Cyber documented in its public research, but the attackers’ sandbox-escape gadget reaches the same code-execution primitive by a different route than the one in the published proof-of-concept.
With this in mind, administrators of self-hosted instances who have not yet applied the July 13th update should do so now.
The security updates also carry Guarded Script, a new feature that restricts the type of code that can run in sandbox contexts, thus making future sandbox escapes less likely.