Nothing says "enterprise software" like a vulnerability that lets a total stranger, with zero login credentials, run their own code on a platform trusted by 85% of the Fortune 500. ServiceNow just had that week, and it did not RSVP for it.

A Sandbox With a Trapdoor

CVE-2026-6875 is a critical remote code execution flaw in ServiceNow's AI Platform, discovered by Searchlight Cyber back in April. Attackers who hit a specific pre-authentication endpoint can escape the platform's script sandbox entirely and execute arbitrary code — no username, no password, no invitation required.

ServiceNow patched its own cloud-hosted instances fast, but self-hosted and partner-hosted customers needed to manually apply a fix released July 13. Over the following weekend, researchers at Defused confirmed attackers were already exploiting it in the wild, using a sandbox-escape technique that reaches the same code-execution outcome by a different route than the original proof-of-concept.

The Patch You Actually Have to Install Yourself

This is the classic self-hosted-software nightmare: the vendor fixes the cloud version instantly, while the on-prem version sits exposed until an admin somewhere gets around to clicking "update." ServiceNow runs over 100 billion workflows a year, which means the blast radius here isn't hypothetical — it's every internal ticketing system, HR workflow, and automation chain built on top of it.

ServiceNow says it hasn't seen exploitation on its own hosted customers, but that's cold comfort if you're running the self-managed version and haven't patched yet. Active exploitation plus a Fortune-500-sized install base is exactly the combo that turns "apply this patch" into "apply this patch today."

If your ServiceNow instance is self-hosted and you haven't touched KB3137947 yet, consider this your sign — before someone else touches your instance for you.

Source: BleepingComputer