If you’re anything like the IT and security leaders I talk to, summer didn’t really slow down, and Q4 planning is already stacking up on your desk. So I’ll keep this one short: a quick recap, some new research from our threat lab, and a stat on AI agents that’s been sticking with me. Three minutes, if you’ve got them.
A Quick Recap, Not a New Announcement
This summer, G2 named Aryaka a Leader again: 16 badges, #1 in SD-WAN across Relationship, Results, and Usability, and a High Performer badge in SASE, based on more than 70 verified reviews and a 4.6 out of 5 rating. I won’t pretend that’s breaking news anymore, but it’s still worth a mention, mostly because it’s not us saying it.
“The Aryaka team ensured the transition was seamless…technical implementation was dedicated and committed.” Verified reviewer, via G2
From our Threat Research Lab: DonutLoader
Unified SASE is only as good as the security intelligence behind it. Our threat researchers just published a technical breakdown of DonutLoader, a multi-stage malware chain that disables Windows’ built-in defenses, hides inside legitimate processes, and reports home over Telegram. It’s the kind of adversary behavior our Threat Research Lab tracks every day, and it’s the same intelligence baked into what we defend against inside Unified SASE.
The Stat That’s Been Sticking With Me
AI agents are shipping faster than most security stacks can track them. New data puts a number on it. Security teams can see only 44% of the AI agents and tools touching sensitive data in their own organizations. 88% of enterprises running agents in production have already had a security incident tied to one, and the average AI agent-related breach now costs $4.7 million.
If your network can’t tell you what an autonomous agent just did with your data, that’s worth closing before you scale AI further, not after.
Sources: Nokod Security, 2026 State of Security in Business-Built Applications and AI Agents Survey; Darktrace, State of AI Cybersecurity 2026; 2026 industry breach-cost compilation.
The DLP Stats Most Teams Miss
Since we’re already talking about AI agents nobody can fully see, here’s a related gap worth naming. Seventy-two percent of enterprises have already adopted generative AI, and most legacy DLP tools were designed before anyone thought to check what happens when an employee pastes a customer contract into ChatGPT or Copilot. The data’s gone before anyone notices, and the average breach now costs $4.99 million.
Sources: McKinsey, State of AI Report; IBM, Cost of a Data Breach.
The real fix isn’t a new feature, it’s inspection that happens inline, in the same path as your network traffic, instead of bolted on as a separate console that’s always a step behind.
Upcoming Live Webinar
Join our live upcoming webinar – Just as a reminder, if you are in the EMEA region, come hear Aryaka experts share practical strategies for improving secure connectivity, optimizing application performance, and navigating China’s complex networking environment.
Ready to talk specifics?
The fastest way to know if Aryaka fits is a short conversation, not another data sheet.
Questions on any of this? Reply directly and I’ll get you to the right person on our team.
As AI continues to reshape enterprise IT, we’re here to help you build a network that’s ready for what’s next.
Best,

Ryan Livesay
SVP, Global Sales
Aryaka Networks