Analysis

AI security engineering Series — Threat Modeling an Email-to-CRM Agent: Follow the Data, Not the Model

Page published

Earliest dated coverage: 16 Nov 2026 · First observed: 17 Nov 2026 · Latest dated coverage: 16 Nov 2026

Coverage timeline

16 Nov 2026medium.com

Single-source analysis — one report is available.

Why it matters

Email-to-CRM threat modeling helps defenders identify where untrusted messages could influence privileged agent actions rather than relying solely on model-level safeguards.

Arsh Riz’s article introduces threat modeling for an email-to-CRM agent by tracing data flows and trust boundaries, organizing security work around threats, controls, and tests. A linked EchoLeak case study documents zero-click data exfiltration through a crafted email in Microsoft 365 Copilot, providing a concrete example of external content crossing AI trust boundaries.