Vulnerability  ·  2026-11-21

ServiceNow Now Assist for CSM Email Response AI Agent: indirect prompt injection

VulnerabilityMedium impactGlobalCVE-2026-18883
ServiceNow addressed an indirect prompt injection issue in the Email Response AI Agent of Now Assist for CSM. The fix was deployed centrally to the agent collection and requires no customer-side platform version or hotfix.
A deployed customer-service AI agent can be manipulated through the content it is asked to process, letting an attacker who controls an inbound email/case influence the agent's replies, actions or data handling — a real-world indirect prompt-injection path into a production agent.
Indirect prompt injection via content the Email Response AI Agent ingests (email/CSM case content). CVSS 4.0 7.1 (AV:N/AC:L/PR:N/UI:P, high integrity impact). The fix was deployed centrally and requires no platform version upgrade.
Now Assist for CSM Email Response AI Agent before AI agent collection v6.1
No platform upgrade required — remediation was deployed centrally to the Email Response AI Agent (Now Assist for CSM); confirm the managed collection is current.
ServiceNow KB KB3235180NVD record
See this in the live feed Explore related AI security and governance findings — updated every morning.
Open the feed →