What AI Gateway actually does
AI Gateway sits between your application and the AI model providers it calls, giving you visibility (logging of requests and responses), response caching to cut cost and latency, and rate limiting - without you having to build any of that yourself. It doesn't handle your application's own user authentication or business logic; those stay entirely your responsibility.
Securing an AI-powered feature
The right protection for an AI endpoint depends on how it's exposed: a public-facing AI feature typically needs a web application firewall, API security and rate limiting; a feature reserved for signed-in customers needs strong authentication instead or as well; an internal-only tool has different priorities again. Think about which of these applies before you launch.
What's included when you request it
Requesting AI Gateway sets up one named gateway with its core free features - it does not automatically enable AI-provider billing, paid credits, model inference, vector search or retrieval add-ons. Those remain deliberate, separate choices you make afterwards.
Protecting your team, not just your app
AI Gateway addresses traffic your own application sends to AI providers. A different concern - your employees or automated agents sending company data to external AI tools you don't control - is handled by Zero Trust, Gateway, DLP and SaaS security controls instead; see our Zero Trust guide for that.
