'Most Severe AI Vulnerability to Date' Hits ServiceNow

According to AppOmni, ServiceNow tacked agentic AI onto a largely unguarded legacy chatbot, exposing customers’ data and connected systems and enabling an attacker with limited information to gain admin-level access. The researcher, Aaron Costello, who is chief of security research at AppOmni, described this as the most severe AI-driven vulnerability uncovered to date.

Authentication flaws included a universal credential used by every third‑party service that connected to the Virtual Agent API, and a lack of required passwords or MFA for user identity, meaning an attacker could impersonate a user with only an email address and a few target-specific details. Costello demonstrated how an attacker could create tickets, manage workflows, and even generate a new admin account, granting persistent access to one of the most sensitive platforms run by any company.

AppOmni reported its findings to ServiceNow on 23 October, and by 30 October the vendor had rotated the universal credential and eliminated the AI agent used to create the admin account. The issue has since been fixed, but the article notes that threat actors could already be inside affected infrastructures, so organisations should conduct thorough cyber-health checks and ensure AI agents are narrowly scoped and reviewed before deployment.