Cybersecurity Research

CrowdStrike Launches AI Gateway to Give Businesses One Control Point for AI Traffic

CrowdStrike · 13 Nov 2026
Key Takeaway Whatever tools you use, decide which AI services are approved, route that use through a single controlled path, and block direct connections to everything else.

AI tools are no longer just answering questions; they are taking actions. According to CrowdStrike, requests to AI models now come from employee assistants, coding agents, software copilots and custom applications, each connecting to model providers with its own credentials, models and settings. That makes visibility alone insufficient, and security teams also need to control that activity.

CrowdStrike's answer is Falcon AI Gateway, now generally available and included with Falcon Guardian. It routes AI traffic through a central gateway that authenticates each request, sends it to the right provider and model, applies usage limits, and applies policy both before the request reaches the provider and again when the response returns. Teams block direct connections to model providers using their existing network controls, so approved workloads must go through the gateway. CrowdStrike notes it does not discover or redirect AI traffic automatically; it only covers workloads teams choose to route through it.

The gateway shares one policy model with the Falcon sensor, but each enforces in its own place. The sensor acts on the endpoint, including local processes, files and commands, while the gateway acts only on routed traffic and cannot stop local processes. Because traffic passes through one point, the AI Kill Switch can cut off an agent's model access via the gateway. Developers connect by changing a base URL and using a gateway key.

Using AI in an advice or accounting practice? Who regulates what, ten controls and a 20-step checklist ->

Summarised by CISO AI from CrowdStrike, written with Claude Sonnet 5.5. We link back to every original so you can read it yourself.