Fortinet has launched a cloud-delivered service that inspects traffic between employees and AI tools through customers’ existing FortiGate firewalls. It adds a practical security layer for organisations trying to govern AI use without deploying a separate stack.
Watch Desk analysis
What happened
The FortiGuard Inline AI Security Service checks prompts sent to large language models, agents and Model Context Protocol tools in real time. Fortinet says it can inspect the content of those interactions and mask sensitive data as it passes through. Customers with stricter privacy or data-sovereignty needs can instead host FortiAIGate on Kubernetes with Nvidia GPUs, keeping prompts and logs in their own environment.
Fortinet also detailed FortiAIGate features for inspecting images, video and code, routing requests according to content, policy and cost, and setting spending limits for teams or individual users. Its expanded data-loss-prevention engine covers 35 kinds of personal and confidential information, the company says. Administrators can enable more than 50 built-in compliance policies, and the gateway logs prompts, outputs and tool calls for auditors.
Why it matters
The change takes security beyond spotting that someone is using an AI service: it aims to inspect what is being sent, including interactions with agents and tools. For organisations already using FortiGate, the cloud-delivered service could offer a route to AI traffic controls without installing a separate security stack. Self-hosting is available for customers who need prompts and logs to stay inside their own environment.
Our read
This is a substantial attempt to bring AI governance into familiar network-security infrastructure, rather than asking organisations to bolt on yet another console. The useful questions are whether inspection works across the messy mix of models, agents and media employees actually use, and how much control customers get over what is blocked or masked. Fortinet’s feature list is concrete; the performance claims still need to meet real-world traffic.
What to watch
- Which AI services, agents and tools the inspection supports in practice.
- How reliably sensitive information is detected and masked.
- What customers need to run the self-hosted Kubernetes option.
Discussion spark: Should AI traffic be inspected and masked by default on company networks, or should employees and teams decide which prompts are sensitive?
Sources and evidence
- Fortinet extends FortiGate firewalls into AI security and adds new FortiAIGate features – SiliconANGLE (26 October 2026, 22:25 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.