ServiceNow’s Now Assist AI platform has a critical vulnerability that can be exploited through default agent configurations, enabling attackers to conduct second-order prompt injection attacks. This flaw permits unauthorized actions, including data theft and privilege escalation, allowing low-privileged users to bypass security controls by leveraging the privileges of powerful administrative agents.
Researchers from Appomni demonstrated successful proof-of-concept attacks, performing Create, Read, Update, and Delete (CRUD) operations on sensitive records. They managed to send external emails containing confidential information while circumventing existing security measures. Organizations using ServiceNow are urged to take immediate actions to limit exposure by enabling supervised execution modes, disabling autonomous overrides, segmenting agent teams, and monitoring agent behavior for suspicious activities. ServiceNow has acknowledged the aligned functionalities and has updated their documentation to enhance awareness of these risks, emphasizing the need for proper auditing of AI agent deployments.
👉 Pročitaj original: Cyber Security News
