Cato Networks Joins the World Economic Forum’s Innovator Communities. Here Is What It Signals for Your Network.

On September 8, Cato Networks announced it has joined the World Economic Forum’s Innovator Communities, the Forum’s bench of high-growth companies invited into the rooms where policy gets shaped. Cato enters through the Forum’s Unicorn community of private companies valued at one billion dollars or more, and will work through the Forum’s Centre for Cybersecurity on initiatives covering cyber resilience, cybercrime, and artificial intelligence (AI). [Source: Cato Networks]

Membership announcements are usually skimmable. This one is worth two minutes, because the work Cato brings to that table is the same work changing how mid-market networks get defended, including the one you run. It is also work we know firsthand: Matrix has been running Cato’s AI security stack in our own production environment as an early adopter, seeing where it excels and where it needed tuning, and next week we are demonstrating it live.

Cato arrives with a year of shipped work

Cato does not show up to that table empty-handed. It spent the last year shipping the AI security stack it now brings to the conversation. It acquired Aim Security in September 2025 and built the result into Cato AI Security, which gives IT teams discovery and control over which AI tools employees actually use and what data moves through them. Then on August 3 it made Agentic Threat Prevention generally available: AI agents that predict how an attacker would likely move through your specific environment and put protections in place before the attack progresses. The production numbers published at launch are the kind no human team matches: an average of 4.5 million traffic signals processed per customer each week, and more than 71,000 targeted restrictions enforced without an analyst touching a ticket. [Source: Network World]

Cato co-founder and chief executive Shlomo Kramer framed the stakes in the announcement: “AI is moving faster than any technology shift I’ve seen before.” An analyst at the research firm IDC made the defender’s version of the same point at the August launch, noting that a full attack can now take less time than investigating a single alert.

How to read news like this

An invitation secures nothing by itself, so read it for signals instead. First, direction: the Forum’s cyber center now has an operator with production AI-security telemetry at the table, which tells you where the standards conversation is heading. Toward visibility and control of AI use, not prohibition. Second, roadmap gravity: vendors sitting in policy rooms tend to build toward what regulators will eventually expect, which ages well for their customers’ compliance posture. Third, staying power: seats like this go to platforms committed to the category for the long haul, which is worth knowing when you are weighing a multi-year secure access service edge (SASE) agreement.

The part that applies to your network this month

Think of it as two clocks: one tracks how fast your people adopt AI, the other tracks how fast attackers do, and you only control the first. The work the Forum is organizing around starts at home, on the clock you own, and it starts with visibility. In a 2026 survey of more than 900 executives and technical practitioners, 80.9 percent of technical teams had AI agents in testing or production, while only 14.4 percent said everything launched with full security approval. [Source: Gravitee] You cannot write a sensible AI policy for usage you cannot see. So pull egress logs for AI traffic nobody sanctioned, check expense reports and browser extensions for tools that arrived without review, and publish a short approved-tool list. Then put guardrails on the data path rather than on your people: cloud access security broker (CASB) and data loss prevention (DLP) controls inside a SASE platform allow the tool while stopping a customer list, a patient record, or a client file from leaving in a prompt.

Four questions for whoever runs your platform

Whether that platform is Cato or anything else, ask these before your next renewal.

1. What context does its prevention engine actually see: identities, devices, traffic, and data activity, or just logs?
2. What controls can it impose on its own?
3. How fast does protection arrive after a new vulnerability drops, measured in minutes or in maintenance windows?
4. How do automated blocks get reviewed, appealed, and expired?

That fourth question is the one almost nobody asks. Machine-speed prevention with no human veto just moves your outage risk from the attacker’s hands to your vendor’s.

We run this stack in production

Matrix has operated Cato AI Security in our own environment as an early adopter, so we have seen where it excels and where it needed tuning. Next week we are showing it instead of writing about it. On September 17 at 10 AM PT, Matrix chief technology officer Jeremy Ness and Cato’s Liron Hakim Bobrov will demonstrate shadow AI discovery, sanctioned versus personal accounts, and live data guardrails from the Matrix production environment, with a short preview of Agentic Threat Prevention at the end. If your team is anywhere on the AI adoption curve, save a seat at Unprompted.