Versa Networks

Most of the 2026 OWASP GenAI Top 10 risks trigger a network event before they cause real damage. That's the part that gets lost in the application-layer conversation. Guardrails are necessary, but a compromised model doesn't respect its own filters. An agent exfiltrating data over DNS doesn't trigger an app-level DLP rule. And a denial-of-wallet attack on a stolen API key doesn't touch the model at all, it just burns your access until someone reads the bill. We went through all 10 OWASP categories, matched each one to the real-world incidents behind it, and mapped them to specific SASE controls: SWG, CASB, DLP, ZTNA, and RBI. The numbers in this one are worth paying attention to: - A stolen Gemini API key hit $82,000 in 48 hours. - The litellm PyPI compromise reached an estimated 2,488 corporate domains in under 40 minutes. - And Georgia Tech research reconstructed 92% of original text from 32-token embeddings, which means a vector database is not a safe place to lose control of your data. We cover the full OWASP list with the network-layer indicators and SASE control mappings for each category. Use the link in comments to learn more.

Most DLP stacks were not designed. They were assembled. SWG from one acquisition. CASB from another. ZTNA bolted on later. The result is policy scattered across disjointed tools, with no single view of how data actually moves, especially through AI workflows. That is the architecture problem at the root of most modern data exposure. And it is fixable. Join us on October 15, as our experts take a practical look at how AI-driven DLP built into a unified SASE platform closes the gaps that siloed tools leave open. Including how to: - Use AI/ML and LLM-powered classifiers to read context, not just patterns - Surface sensitive data in documents, images, metadata, and compressed files with OCR and multi-modal analysis - Enforce one DLP policy across internet access, private apps, and GenAI Register using the link in the comments.

Plug a laptop into a conference room port. What does your Zero Trust program check next? For many enterprises, the answer is: not much. Remote access and the WAN edge were the logical first phases. Each had a single point of entry to secure and the campus is different. Every access port is its own control point, and VLANs plus port authentication made the problem look solved. They answer whether a device can join the network and they don't answer what it should be allowed to reach once it's there. Our latest blog explains why the flat network persists and how Versa Secure SD-LAN extends Zero Trust and microsegmentation to the LAN edge using the policy model you already run. Read the blog using the link in the comments.

This is the architecture that replaces five consoles. A plain-language request from a NOC analyst hits an enterprise routing agent. That agent decomposes the question and fans it out to the right MCP servers: WAN and security through Versa, switching through network infrastructure, wireless through access, endpoint through the client layer. Each server queries its own system of record. The routing agent correlates the results and one answer comes back. The routing and governance layer is the part most enterprises haven't figured out yet and that's where the operational leverage is. Read the blog post to see the full architecture and how Zero Trust governance wraps the whole thing. Use the link in the comments to learn more.

Recently we were back at Second Harvest of Silicon Valley, packing fresh produce for families across Santa Clara County. Second Harvest quietly does some of the most consistent, high-volume community work in the region, and they make it easy for groups like ours to show up and be useful. There is something clarifying about work where the output is immediate and the people it reaches is impactful. Thank you for having us Second Harvest of Silicon Valley.

Hour zero: a vulnerability is disclosed. Hours later: it's being exploited in the wild. Eight months later: the average organization has finally identified and contained the breach. That is not a gap you close by adding another tool to a stack of 45. The math only changes when the architecture changes: continuous visibility, inline enforcement, and Zero Trust applied at every point of interaction, in every direction. This is the problem Pervasive Security was built to solve. Read Part 1 of our series using the link in the comments.

Enterprises are ready to consolidate SASE, but the hesitation is operational. Migration complexity leads consolidation concerns at 51%, with cost of transition and performance limitations follow closely, and internal organizational resistance ranks last at 12.5%. That gap indicates that the appetite is there, but the teams need confidence to adopt without disrupting what is already running. Futuriom surveyed 144 CxO and director-level IT, security, and networking executives to capture how peers at organizations like yours are thinking about SASE platform consolidation right now. The survey included the risks, the benefits, the tradeoffs, and what success looks like over the next one to three years. Use the link in the comments to access the complimentary report.

Most AI-NetOps deployments stall before they deliver value. Not because the AI is wrong, but because engineers cannot validate whether it is right. Gartner put it plainly: adoption of AI in NetOps depends on explainability of outcomes, traceable network evidence, and contextual awareness of telemetry. Those are architectural properties and you either have them or you do not. We built this trust layer post to walk through exactly this: - Why cross-domain telemetry correlation matters - How our Zero Trust MCP Server changes the agentic AI governance picture - What a phased NetOps roadmap actually looks like for teams that need to show ROI before going hands-free The 45% MTTR reduction customers are reporting after deploying the MCP Server is a good number. The reasoning behind it is what makes it stick. Read the post using the link in comments.

AI is changing how threats move and it's changing how security platforms need to respond. We'll be at Les Assises this week, walking through how AI is embedded inside Universal SASE and what that looks like in practice for network and security teams. Visit us at booth H33. We look forward to connecting with you.