Vendors & EquipmentCato NetworksAgentsenseSaseAI Security
Cato Networks launches AgentSense, calls zero trust 'not enough' for AI
Cato Networks ships AgentSense, tracing AI agent behavior at runtime, blocks prompt-injection hijacks, free with AI Usage licenses, built on its $320M Aim Security deal.
Vendors & EquipmentWhy it matters
- Cato AgentSense is included at no additional charge within Cato AI Usage licenses.
- The technology derives from Cato's $320 million acquisition of AI security firm Aim Security.
- Cited threats include EchoLeak in Microsoft 365 Copilot, a zero-click data theft vulnerability.
- Cato CEO Shlomo Kramer calls agent security 'a completely new security stack'.
The story
Cato Networks has folded a real-time AI agent monitoring capability into its SASE platform and included it at no additional charge within its Cato AI Usage licenses. The product, Cato AgentSense, traces how instructions and data influence an AI agent's behavior at runtime and flags any deviation from the agent's original mandate.
The launch comes with a blunt message: zero trust is no longer sufficient in the AI era. The SASE vendor argues that identity-based access controls, the foundation of the zero trust model, cannot police autonomous software that acts on prompts and external content after access has already been granted.
Why does Cato think zero trust falls short?
AgentSense is designed to close that gap. It monitors the decision chain inside an AI agent, detects when untrusted content steers the agent toward actions it was never authorized to take, and blocks those actions before execution. The primary threat model is indirect prompt injection, where an attacker plants malicious instructions in content the agent legitimately processes.
Cato pointed to three disclosed attack vectors as evidence the risk is practical rather than theoretical:
- EchoLeak, a zero-click vulnerability in Microsoft 365 Copilot that let attackers silently extract sensitive organizational data with no action required from the victim.
- CurXecute, a remote code execution flaw disclosed in 2025 that targets the Cursor AI code editor through its model context protocol (MCP) integration.
- DuneSlide, a set of flaws allowing prompt-injected instructions to bypass Cursor's built-in sandbox protections and trust boundaries.
"These examples underscore why enterprises need to trust that an AI agent will carry out the task they authorized, even when it encounters content an attacker controls," said Itay Ravia, head of Cato AI Labs. "AgentSense helps preserve that trust by tracing how untrusted content influences an agent's decisions and sensitive actions, helping enterprises adopt AI agents without giving attackers control of the workflow."
Where did the technology come from?
The service is not organic engineering. It derives from Cato's $320 million acquisition of Aim Security, an Israeli firm specializing in securing enterprise deployments of large language models and generative AI. That deal gave Cato the runtime tracing and agent-assessment technology now embedded in its flagship SASE platform, which positions the company to sell AI security as an extension of an existing network contract rather than a separate product line.
What does this change for enterprises already running Cato SASE? Effectively, agent monitoring arrives as a licensing line item rather than a procurement exercise, bundled into AI Usage licenses the vendor already sells.
A new security stack, in the CEO's words
Cato CEO Shlomo Kramer framed the product as more than a feature addition. Today's security stacks, he said, need "to cover the prompts and responses and action of agents across the muscle of the organization. It has to build the whole lifecycle of security, from posture management to protection to detection response, and back to posture."
"This is a completely new security stack that is very different than the network traffic and the application and data security stacks," Kramer said.
That claim carries commercial weight. Cato competes in a SASE market where rivals Cisco, Palo Alto Networks, Zscaler and Netskope are racing to attach AI controls to their platforms, and a security stack purpose-built for agents is a differentiating pitch. Buyers should note the distinction: runtime agent tracing is a genuine technical advance over static access policy, but "completely new stack" remains a vendor framing until independent validation of AgentSense's detection coverage exists.
Cato has not disclosed pricing tiers for AI Usage licenses or a phased rollout schedule beyond immediate availability within the platform. Enterprises weighing adoption will find the decision point straightforward: agent runtime monitoring is now part of the SASE bill, and competitors without an equivalent capability will face pressure to answer it.
Also reported
Source: sdxcentral.com
More from James Calloway
Show full bio
Staff writer covering companies, deals and strategy in Carrier business, 5G and 6G, fiber and broadband at Telecom Gazette.
545 articles