AI is creating an emerging security problem that has less to do with smarter AI-fueled attacks and more to do with internal traffic patterns organizations were never designed to monitor. The byproduct is a new attack surface hiding in plain sight.
AI is driving the emergence of new types of enterprise traffic. Traditionally there is human behavior traffic and application telemetry traffic. Now AI assistants, agents, and automated systems generate a third category that deserves to be treated separately, writes Lori MacVittie, distinguished engineer and chief evangelist at F5, in a recent post.
For decades, most enterprise traffic followed familiar patterns. People opened applications, clicked websites, downloaded files, sent emails, and connected to databases, she writes. Applications talked to other applications, but the interactions were generally predictable enough that security teams could establish a baseline of normal behavior.
A worker might open an AI assistant and ask for a sales summary, customer update, or project status report. What looks like one request from a person may actually be an AI system retrieving data from a CRM, pulling documents from a cloud repository, querying internal databases, checking calendar information, and firing off requests to one or more large language models — all within seconds, before the agent responds.
“Automation traffic consists of machine-generated actions interacting with infrastructure. Agents call APIs. Automation platforms update configuration. Decision systems trigger workflows. These requests are no longer initiated primarily by humans. They are generated by systems reacting to signals and context,” MacVittie writes.
AI systems complicate traditional assumptions about user and application traffic. Security teams now have to understand traffic that behaves like people but originates from software.
MacVittie says this starts sounding less like a traditional network monitoring challenge and more like an identity problem.
“Identity and authorization controls must ensure agents operate only within their intended scope,” MacVittie writes.
Traffic itself becomes the next security signal.
Machine-driven traffic can be spoofed, replayed, or buried inside legitimate automation flows, and forged telemetry can quietly steer the AI systems consuming it. Without inspection at this layer, an adversary inside the network operates at machine speed with machine-level authority — and security teams lose the visibility they need to know it’s happening.
The danger isn’t only an attacker moving laterally — it’s an AI control loop reacting, at speed, to signals an attacker has shaped.
“Automation produces automation traffic. Observability produces telemetry traffic,” MacVittie writes.
Both are natural byproducts of AI at scale, MacVittie argues, and both need to be treated as first-class traffic — authenticated, governed, and secured like anything else on the network.
Photo by Denys Nevozhai on Unsplash
