Your AI Agents Are Already Talking to Each Other. Who’s Watching? – Netskope Triple-T
With the average enterprise seeing users access over 60 unmanaged AI apps and agents, agentic traffic has rapidly become a ubiquitous—and largely ungoverned—part of the modern network. Most zero trust frameworks in production today assume a human is behind every request. Agentic AI breaks that assumption, rendering traditional detection and standard DLP insufficient. True agentic governance demands machine-native zero trust that enforces agent identity, maintains continuous provenance tracking, and establishes runtime policy enforcement at the precise point of decision-making. Beyond simple monitoring, organizations need full explainability and auditability into agent choices, ensuring trust is earned and verified at every step of execution.
In this session, Netskope will lay out what a machine-native zero trust architecture requires: agent identity as a native participant, policy enforcement at the moment of inference, and data lineage that survives copies, reformats, and RAG pipelines. We’ll walk through concrete guardrail patterns for stopping prompt injection and data leakage before they reach production, and how data lineage turns “we think it’s fine” into “here’s the record.”
Learn how to architect for an enterprise where agents outnumber employees, and trust is earned one transaction at a time.
Mark Fabbi is a 40-year networking veteran and strategic advisor. For 28 years, he was Vice
President Distinguished Analyst for Gartner, driving networking research. He defined the
application delivery controller market (ADC) and was a key influencer in the WAN Optimizations
space. Mark’s research foretold the migration of network and security from data center devices
to cloud services, predating SASE. Before Gartner, he helped bring emerging tech to
enterprises and created one of North America’s first managed network service offerings. He
currently serves on the Netskope CxO Advisory Board.
