What happened and why it matters
ServiceNow just pushed security updates for three AI Platform vulnerabilities—CVE‑2026‑18885, CVE‑2026‑18886 and CVE‑2026‑74820—that can be triggered without any credentials or user interaction.
At the same time the vendor fixed a high‑severity sandbox‑escape issue, CVE‑2026‑6876, that also lives in the same platform.
Technical specifics
The three newly‑rated maximum‑severity flaws break down as follows:
- CVE‑2026‑18885: classic code injection allowing an attacker to run arbitrary commands on the server.
- CVE‑2026‑18886: privilege‑escalation code injection that lets a low‑privileged user jump to admin‑level rights.
- CVE‑2026‑74820: SQL injection that can expose or corrupt the underlying database.
All three are exploitable in low‑complexity attacks, require no prior authentication, and do not need a victim to click anything.
The sandbox‑escape CVE‑2026‑6876 lets a malicious actor break out of ServiceNow’s isolation mechanisms, potentially reaching the host system.
Who’s at risk
Any organization running the ServiceNow AI Platform—whether on the Xanadu, Yokohama, Zurich or Australia release branches—is exposed until the patches are applied.
ServiceNow says it has not seen any active exploitation of the three new flaws, but the vendor’s own advisory notes that the attack surface is trivial.
Two years ago, a chain of CVE‑2024‑4879, CVE‑2024‑5178 and CVE‑2024‑5217 was used to breach private firms and government agencies, showing how quickly multiple bugs can be combined for a larger compromise.
In July, Defused reported that CVE‑2026‑6875, a pre‑auth sandbox escape in the same platform, was being weaponised in the wild.
What to do now
- Apply the latest ServiceNow patches to every release branch you operate.
- Verify that the patches are active by checking the version numbers against ServiceNow’s advisory.
- Audit API endpoints for any unexpected data exposure; the vendor disclosed a recent incident where researchers queried customer instance data through an unauthenticated flaw.
- Consider temporary network segmentation for the AI Platform until you confirm the updates are in place.
- Monitor security feeds for any emerging exploits targeting CVE‑2026‑6875 or the newly patched CVEs.