Cato Networks Unveils AgentSense to Stop Harmful AI Agent Actions Before Execution

What happened
Cato Networks introduced AgentSense, a new security feature aimed at stopping harmful actions by AI agents before they are carried out. AgentSense works by monitoring how untrusted content influences an agent’s decisions during its operation. The tool is now included at no extra cost for customers with Cato AI Usage licenses. Alongside the launch, Cato Networks announced a $30,000 challenge to test the effectiveness of AgentSense’s protections.
Key details
- AgentSense is developed by Cato AI Labs and integrated into Cato’s platform.
- It is available for free with Cato AI Usage licenses.
- The technology tracks the influence of untrusted content on AI agent actions in real time.
- A public challenge offers $10,000 each to the first three eligible participants who can bypass AgentSense’s protections.
- AgentSense aims to prevent indirect prompt injection attacks within authorized workflows.
Why it matters
Organizations using AI agents to automate tasks face risks from attackers who try to manipulate agent behavior through external content. With AgentSense, companies using Cato’s platform gain a new layer of protection that can block unsafe actions even when agents have legitimate access. Security professionals now have an opportunity to test and potentially improve these protections through the public challenge, which could influence how enterprises secure AI-driven workflows.
