● Product: FortiGuard Inline AI Security Service, on FortiGate
● Inspects: LLM prompts and Model Context Protocol traffic, in real time
● Controls: data masking, prompt blocking, spending caps
● Policies: more than 50 built in
● Self-hosted: Kubernetes with Nvidia GPUs
Fortinet has extended FortiGate firewalls with the FortiGuard Inline AI Security Service, which uses FortiAIGate to inspect AI traffic in real time - including prompts sent to LLMs and Model Context Protocol traffic. It masks sensitive data, blocks prompts, and ships more than 50 built-in compliance policies.
MCP inspection at the network layer is the notable part. MCP is how agents reach tools and data, and until now it has largely been an unmonitored channel inside organisations that have otherwise locked down egress. A firewall that reads it is a different security posture from one that sees an encrypted call to an API.
Administrators can also set spending caps, which addresses the other quiet problem: an agent in a loop is a billing event before it is a security event. Self-hosted deployments run on Kubernetes with Nvidia GPUs.
Sources
Rio Rundown, 27 October.