ServiceNow AI Platform sandbox escape lets admins run code on host
A flaw let a logged-in administrator break out of the sandbox that limits scripts, and run commands directly on the underlying server. ServiceNow has already fixed it on hosted instances.
| Severity | CriticalCVSS 9.4 of 10 |
|---|---|
| Being exploited? | Not knownNo attacks reported so far |
| Fix available? | YesUpdate applied to ServiceNow hosted instances; update provided to partners and self-hosted customers. Specific patched version numbers not stated in the material. · since 29 Oct 2026 |
| How sure is this? | Official sourceBacked by an official source |
How the attack works simplified
- 1Admin accountAttacker holds valid admin credentials
- 2Sandbox bypassEscapes Rhino sandbox restrictions
- 3Host code executionRuns arbitrary code on the host OS
- 4Further accessReaches beyond the platform
- ResultAttacker runs code on the host server
What happened
A sandbox escape in the ServiceNow AI Platform allows an authenticated user with the admin role to bypass Rhino sandbox restrictions and execute arbitrary code on the host operating system.
Are you affected?
You are affected if you run the ServiceNow AI Platform on a version prior to the fix, including self-hosted or partner-managed instances, and have users with the admin role.
ServiceNow AI Platform versions prior to the remediated release (exact versions unknown from the material)
What to do
- Check with ServiceNow or your partner whether your instance has the update, and confirm the patched release number
- Apply the update or upgrade to a patched release if you run a self-hosted or partner-managed instance
- Review which accounts hold the admin role and remove any that are not needed
Timeline
- 29 Oct 2026, 19:17 UTCPublicly disclosed
- 29 Oct 2026, 19:17 UTCReported by NVD · NVD
- 29 Oct 2026, 19:21 UTCPicked up by Zero Radar
Sources 1 report, oldest first
Official vendor, government CERT or advisory database
- CVE-2026-103093NVD · 29 Oct 2026, 19:17 UTC
Technical details
- ID
- CVE-2026-103093
- CVSS
- 9.4 / 10
- Disclosed
- 29 Oct 2026, 19:17 UTC
- On Zero Radar
- 29 Oct 2026, 19:21 UTC (4 min after disclosure)
- Last change
- 29 Oct 2026, 19:21 UTC
The plain-language parts and the attack diagram were written by an AI model (anthropic/claude-haiku-5.5) from the sources above. Facts like versions, exploitation and dates come from the sources themselves. How Zero Radar works