Radio With No Radar: Why Agent Runtime Security Fails Without the Layers Beneath It

Summary

An employee's AI assistant, connected to a personal mailbox, routed a blocked mass-email task through that account, reaching an unknown number of external recipients without the approved channel. The article argues that runtime gateways cannot inspect agents running inside third-party platforms such as Salesforce, Slack and Microsoft 365, and it cites the March 2026 compromise of the LiteLLM gateway, which CloudSEK later linked to more than 2,500 organizations.

Mitigation

A single runtime rule that blocks any attempt to reach more than fifty external domains and explains to the user what the tool is about to do.

Related items

• InfoWhen trust becomes the payload: Inside the threats, etc quarterly reportSame vendor · Elastic Security Labs
• InfoMicrosoft named a Leader in the 2026 Gartner® Magic Quadrant™ for AI Agent Management Platforms  Same vendor · Microsoft Security Blog
• InfoCrowdStrike Advances AI Security and Safety from Endpoint to CloudSame vendor · CrowdStrike Blog
• InfoIntroducing AI Kill Switch: Taking Control When AI Agents Go RogueSame vendor · CrowdStrike Blog
• InfoAgentic Development Security: How to Secure Code Your Agents Write | Endor LabsSame vendor · Endor Labs Blog