The Network Is The First Line Of Defense – It’s Time To Treat It That Way
Editor’s Note: This article was originally published on Forbes Business Council on May 21, 2026. It is republished here following the conclusion of the platform’s exclusivity period.
For the past decade, enterprise security has operated on a reasonable assumption: Employees connect to applications in the cloud, so inspect and enforce everything on the way out, and assume most threats originate outside the organization.
This assumption is no longer sufficient. And the gap between what traditional architectures were built to handle and what enterprises face today is widening fast.
Artificial intelligence is fundamentally an inflection point. It is not just that there is more traffic or that it moves faster. The nature of the traffic itself has changed.
Instead of just securing people or devices connecting to applications, today there is a rapidly increasing volume of AI elements: LLMs, agents, AI workloads. Digital sovereignty and the business case are driving these elements to move from outside the enterprise perimeter to everywhere.
Why? AI elements such as agents, MCP servers, LLMs and AI workloads need to dynamically and constantly connect to other AI elements. Therefore, the flows of traffic are now shifting from inside-out to outside-in, inside-to-inside, essentially anywhere and everywhere. Moreover, the accessibility needs of these elements are expanding the attack surface significantly.
Security architectures designed around outbound inspection and web filtering were not built for this. Many enterprises today use a combination of cloud-based security for in-to-out traffic and firewalls for out-to-in protection, creating a fragmented approach and increasing security risk posture.
In this environment, the network itself must become the enforcement layer, what I call an enforcement fabric that applies security policy and intelligence across every edge and every traffic flow.
A Different Architecture
In my experience, the enterprises best positioned for the AI era are the ones that rethink the network itself as an enforcement layer. The network already touches everything—every user, every device, every application, every workload. That puts the network in a unique position: It can see everything, and it can act on what it sees in real time.
What that means in practice is building an enforcement fabric: a shared intelligence and policy layer across every edge, every network segment and every traffic flow. Rather than treating networking and security as separate disciplines integrated after the fact, the enforcement fabric should treat them as a single operational discipline from the start.
When networking and security are genuinely unified, the network becomes more than a transport layer. It becomes the connective tissue of a secure digital business.
Policy should be applied consistently across branches, campuses, data centers, cloud environments and OT networks. Contextual intelligence should be shared across systems in real time. And security operations platforms should trigger remediation programmatically, at machine speed, rather than waiting for an analyst to review a log file hours after an event occurred.
Why Directionality Matters
One of the most important and least discussed limitations of current SASE architectures is that many are built for inside-out traffic, inspecting connections initiated from inside the organization going out. Protection from outside-in traffic still requires firewalls at each perimeter connected to the internet.
In the AI era, this is a critical gap.
AI agents do not initiate communications in one predictable direction. They communicate across environments, access internal systems, interact with external services and spawn additional automated processes. An autonomous agent may initiate a connection to an internal database. An application may call another application across a cloud boundary. A workload may move laterally across distributed infrastructure without any human initiating the exchange.
Securing that kind of traffic requires enforcement that operates in every direction. An enforcement fabric should extend zero-trust controls and inspection across all flows, regardless of origin or destination, without requiring separate infrastructure or additional firewalls at every location.
The Blueprint
For CIOs, CISOs and CEOs thinking about where to invest in the next phase of infrastructure, the question is not which security products to add. The question is whether your architecture is capable of becoming the enforcement layer of your entire digital business.
That means unifying networking and security under a single policy framework. It means building enforcement into every layer of the network, not just the edge. It means designing for multidirectional traffic flows, not just outbound web access. And it means creating the conditions for security operations to move from alert-and-respond to detect-and-act.
I think the enterprises that will be most resilient in the AI era are the ones that stop treating security as something layered on top of infrastructure and start treating the infrastructure itself as the enforcement layer.
That is the network of the future. The time to build toward it is now.
Kelly Ahuja is a seasoned industry veteran with more than 20 years of experience in networking and telecommunications. He currently serves on the board of directors for two startups in Silicon Valley.