Agent runtime security fails for agents that bypass enterprise gateways

The article argues that runtime enforcement for AI agents depends on traffic passing through an enterprise-owned gateway, so most agents fall outside real-time blocking controls. It cites a recent incident in which an employee used an AI assistant to trigger a mass-email automation. The excerpt is truncated and names no CVEs, affected products or versions. Defenders should review which agent traffic avoids their gateways and consider controls at the layers beneath the runtime, such as the automations and permissions agents can invoke.

For security analysts · Written by AI from 1 source · Wrong?