ServiceNow agentic AI attack — Jan 2026

What happened

In January 2026, a security incident involving the ServiceNow Virtual Agent highlighted how traditional authentication and authorization failures can be amplified within agentic systems. The incident resulted in a platform-wide compromise, demonstrating the risks associated with granting autonomous agents excessive permissions without robust identity management.

According to the report, the compromise was driven by three cascading failures:

• Hardcoded credentials: The system utilized a shared token across all customer environments, creating a single point of failure.
• Broken identity verification: The platform accepted email addresses as sufficient proof of identity, bypassing necessary multi-factor authentication (MFA) or single sign-on (SSO) requirements.
• Excessive agent privileges: The Virtual Agent was granted broad permissions, allowing it to create data anywhere within the platform, including the ability to modify or create administrative accounts.

While the AI agent did not introduce new vulnerability types, its autonomous nature allowed it to chain these existing security flaws together, leading to a full platform compromise. This incident serves as a reminder that when agents are empowered to perform actions on behalf of users, failures in foundational security controls can lead to significantly more severe outcomes than in traditional, non-agentic applications.