# ServiceNow: an email can steer the Now Assist CSM reply agent

> CVE-2026-18883 is an indirect prompt injection in ServiceNow's Email Response AI Agent, fixed centrally by ServiceNow in November 2026.

- **Severity:** Medium: Real and reproducible, but needs unusual setup, user help or a narrow audience. Plan a fix.
- **Category:** Vulnerability
- **Published by:** ServiceNow
- **Disclosed:** 2026-11-19
- **CVE:** CVE-2026-18883
- **CVSS:** 7.1
- **Affects:** ServiceNow Now Assist for CSM, ServiceNow AI Platform
- **Research by:** ServiceNow
- **Primary source:** https://github.com/advisories/GHSA-f7fm-26fm-r3gq
- **Page:** https://theinjection.dev/items/servicenow-now-assist-email-agent-prompt-injection/

ServiceNow published CVE-2026-18883 on 19 November 2026, an indirect prompt injection in the Email Response AI Agent in Now Assist for CSM. The fix went out centrally, so customers do not need to upgrade.

## What happened

ServiceNow published CVE-2026-18883 on 19 November 2026 for an indirect prompt injection in the Email Response AI Agent in Now Assist for CSM. The company scores it 7.1 on CVSS v4 and says it is not currently aware of malicious exploitation. A second AI flaw, CVE-2026-125288, was published the same day.

## How it works

Indirect prompt injection places attacker instructions in content an agent reads instead of in a prompt a user types. The CVSS v4 vector ServiceNow assigned to CVE-2026-18883 needs no privileges and only passive user interaction, and rates integrity impact high with confidentiality low, which matches instructions arriving inside a message the agent processes on its own. ServiceNow publishes no payload and no indicators.

## Who is affected

Customers running the Email Response AI Agent in Now Assist for CSM are in scope for the prompt injection. The second flaw, CVE-2026-125288, is a code injection in the ServiceNow AI Platform that an authenticated user with template editing privileges could use to run code in another user's session. Affected version ranges are not published for either.

## What to do

Nothing to install for CVE-2026-18883: ServiceNow states the fix was deployed centrally to the agent, with no platform upgrade, hotfix or version backport required. For CVE-2026-125288, ServiceNow deployed the fix to hosted instances and supplied it to partners and self-hosted customers, and recommends applying the update or upgrading to a patched release promptly.

## Quick facts

- **CVE:** CVE-2026-18883 prompt injection, CVE-2026-125288 code injection
- **Affected:** Email Response AI Agent in Now Assist for CSM; ServiceNow AI Platform
- **CVSS:** 7.1 and 8.4, CVSS v4, scored by ServiceNow
- **Status:** Fixed; CVE-2026-18883 deployed centrally to the agent
- **Exploitation:** ServiceNow says it is not aware of malicious exploitation
- **Disclosed:** 19 November 2026

## Questions

### Do I need to upgrade ServiceNow to fix CVE-2026-18883?

No upgrade is required for CVE-2026-18883. ServiceNow states that the fix was deployed centrally to the Email Response AI Agent, with no platform upgrade or hotfix needed and no version-specific backports. CVE-2026-125288, the code injection published the same day, is different: ServiceNow deployed it to hosted instances and supplied the fix to partners and self-hosted customers to apply.

### Which ServiceNow versions are affected by CVE-2026-18883?

ServiceNow publishes no affected version range for CVE-2026-18883. The advisory names the Email Response AI Agent in Now Assist for CSM as the affected component and lists both affected and patched versions as unknown. Because the remediation was delivered to the agent rather than to the platform, there is no release number for a customer to check.

### Has CVE-2026-18883 been exploited in the wild?

ServiceNow says it is not currently aware of malicious exploitation of CVE-2026-18883. The company scores the flaw 7.1 on CVSS v4, with a network attack vector, no privileges required and passive user interaction, and rates the integrity impact high and confidentiality low. No payload details or indicators of compromise are published with the advisory.

## Sources

1. [ServiceNow Advisory](https://github.com/advisories/GHSA-f7fm-26fm-r3gq)
2. [Code Injection CVE](https://github.com/advisories/GHSA-jq2f-6vj2-4r5g)
