Sharon Halbani

Can your organization adopt AI at speed without risk outrunning control?  ⚠️ Every AI use case changes how data is accessed, used, and protected. At SASEfy 2026, you’ll hear from Microsoft, Forrester, Dayforce, and Cato on where AI risk lives, how agentic AI impacts Zero Trust, and what good looks like in AI security. Plus, spin to win F1 merch! Save your spot now: https://bit.ly/4uGz3TN

5

‏1‏ ‏תגובה‏

As Claude continues to become an integral part of many enterprise AI strategies, Cato AI Security now supports Claude’s Compliance API, giving organizations better visibility into Claude Enterprise activity while keeping AI governance in one place. A nice example of making enterprise AI adoption easier. Learn more: https://bit.ly/4wPDWuM

11

What do people love more than secrets? Deciphering them.🕵️‍♂️ Join Etay Maor, Cato CTRL’s founding member and chief security strategist at Cato Networks, as he breaks down the real code behind AI prompt injections. Watch the full session: https://bit.ly/4r4RC35 #cybersecurity #AIsecurity

9

Identity and access controls alone don't cut it when local coding agents can execute risky actions in real time. To truly protect your environment, you need an approach that combines visibility (what exists in my environment), posture (what risk does it carry), and runtime control (stopping risky actions before they happen). My latest blog post breaks down Cato Networks's 3-stage framework for agentic AI security. Check it out below:

6

Today, I had the pleasure of presenting at Cato Networks AI Day. My talk focused on AI efficiency, and I’d like to share three key takeaways that anyone can apply today. 1. Context is king Treat context with respect—input tokens and cache tokens are often the dominant part of your bill. By being intentional about what you send to the model, you can significantly improve both cost efficiency and performance. 2. Choose the right model for the task It might sound obvious, but many organizations default to always using the newest and most powerful model. In many cases, upgrading models can add up to ~40% in cost. Before switching, ask yourself: is the performance gain worth the premium? Run proper evaluations and benchmarking on your specific use case to ensure you’re upgrading for the right reasons. 3. Always check your default settings Most tools come with strong default configurations. For example, in many models, “thinking” (or reasoning capabilities) is turned on by default, and the effort level is often set to high. Before starting a session, make it a habit to review your settings to ensure they are the right fit for your task. Simple principles, but highly effective when applied consistently.

66

‏8‏ ‏תגובות‏

Cato has achieved AWS Security Competency in AI Security status in the Agentic AI category. For our product teams, the really interesting part is what we had to build to earn that badge. Cato AI Security runs on AWS and helps secure AI built on AWS, including environments using Amazon Bedrock and Amazon SageMaker, while keeping AI security connected to the same platform securing the rest of the enterprise. That's how we think about architecture at Cato: build the pieces to work as one system instead of asking customers to stitch them together later. Proud of the teams who keep pushing that architecture forward. https://lnkd.in/deSa_ysv

63

We are at a Mythos moment ⚡ Advanced models like Mythos may accelerate attack workflows, but they can also create new opportunities for defenders. The real differentiator will not be the model alone, but the scaffolding around it: the visibility, telemetry, workflows, and inline enforcement that turn model intelligence into defensive action. I collaborated with my colleague Avishay Zawoznik at Cato Networks on this post, where we explore both the risks and the longer-term defensive potential of these models. Read the full blog post: https://lnkd.in/epiDeQQg

65

Agentic attacks do not expose a lack of security tools. They expose what happens when those tools depend on human-speed workflows, disconnected context, and manual response. Frontier AI models like Mythos and GPT-5.5-Cyber are changing the economics of cyberattacks. They can compress the attack lifecycle, lower the skill required to execute sophisticated activity, and enable threats that automate and adapt. That creates a different kind of problem for security teams. In this 15-minute webinar, Cato Networks CEO Shlomo Kramer and VP of Threat Intelligence Etay Maor will discuss how agentic attacks are evolving, why fragmented architectures fall behind, and what security leaders need to understand now. The attack model is changing. Security architecture has to change with it. Register now: https://bit.ly/4erFaos

18

Agentic attacks do not expose a lack of security tools. They expose what happens when those tools depend on human-speed workflows, disconnected context, and manual response. Frontier AI models like Mythos and GPT-5.5-Cyber are changing the economics of cyberattacks. They can compress the attack lifecycle, lower the skill required to execute sophisticated activity, and enable threats that automate and adapt. That creates a different kind of problem for security teams. In this 15-minute webinar, Cato Networks CEO Shlomo Kramer and VP of Threat Intelligence Etay Maor will discuss how agentic attacks are evolving, why fragmented architectures fall behind, and what security leaders need to understand now. The attack model is changing. Security architecture has to change with it. Register now: https://bit.ly/4esH8q5

16