HiddenLayer integration with Truefoundry AI Gateway

Every team shipping AI right now is living the same contradiction. Developers want to wire up new models as soon as they appear. Security teams need every prompt and every response to be safe, compliant, and free of leaked data. Both are right, and neither can wait for the other. So the real question is not which model you secure. It is how you enforce a single, consistent security policy across the dozens or hundreds of models your teams are already calling, without rebuilding the integration every time.

That gap, between accessing models and actually governing what flows through them, is where most AI security efforts quietly fall apart. Today, we are closing it. We are announcing the integration of HiddenLayer Runtime Security with the TrueFoundry AI Gateway, so you can apply runtime AI security to any model behind the gateway from one place, with no per-model engineering.

The TrueFoundry AI Gateway

The TrueFoundry AI Gateway is the control point for enterprise AI. It gives developers unified access to hundreds of LLMs through a single API, with smart routing, observability, and cost controls already built in. As adoption scales, the hard part is no longer reaching a model. It is managing everything that comes after: multiple providers, shifting APIs, and compliance requirements that apply to every call.

The gateway turns that sprawl into one consistent surface. Every request and every response passes through the same path, which means security no longer has to be bolted onto each application separately. It can be enforced once, centrally, for everyone.

HiddenLayer Runtime Security for AI

HiddenLayer Runtime Security is a runtime security platform built specifically for LLM applications. Instead of trusting that a prompt or a response is safe, it inspects each interaction as it happens and returns a clear decision.

It catches prompt injection and jailbreak attempts before they reach the model, so an attacker cannot trick the system into revealing its instructions or bypassing its guardrails. It detects and redacts sensitive data, such as email addresses and phone numbers, so confidential information never silently slips into a prompt or out in a response. And because policies live in the AISec Platform Console rather than in your code, your security team can tune what is blocked, what is redacted, and what is merely flagged, without anyone shipping a new release.

Better Together: One Integration, Every Model

On their own, each tool is strong. Together, they remove the work that usually makes AI security painful. Instead of embedding an SDK into every application or maintaining separate security code for each provider, you connect HiddenLayer to the gateway once. From that point on, any model routed through TrueFoundry automatically inherits the same protection.

The gateway listens for HiddenLayer's verdict on every call and enforces it on the spot. A malicious prompt is blocked. Sensitive content is redacted in place, and the cleaned text moves on. A safe request passes straight through. Your developers keep moving at full speed and barely notice the layer is there, while your security team gets centralized policy control and a complete audit trail spanning every model and environment.

How the Integration Works

HiddenLayer runs as a lightweight wrapper that the gateway calls on the way in and on the way out. Each request makes a quick, predictable round trip before it ever reaches a model.

‍

A layered path for every request

1. Your application sends a request, and it first arrives at the TrueFoundry AI Gateway.
2. The gateway forwards the prompt to HiddenLayer, where it is evaluated against your configured security policies for threats such as prompt injection, jailbreak attempts, and sensitive data exposure.
3. HiddenLayer returns a verdict to the gateway — Allow, Block, or Redact. Blocked requests are stopped, while flagged content can be automatically redacted before proceeding.
4. If the request is allowed, the TrueFoundry AI Gateway routes it to the target model provider selected by your application.
5. The model generates a response and returns it to the TrueFoundry AI Gateway.
6. The gateway applies the policy decision received from HiddenLayer and prepares the final safe response.
7. The secured response is delivered back to your application.

This architecture enables organizations to enforce consistent AI security policies before model execution, helping prevent prompt injection attacks, jailbreak attempts, and sensitive data exposure while maintaining a unified access layer for all AI models.

Get Started

AI security should not slow you down. With HiddenLayer Runtime Security on the TrueFoundry AI Gateway, you get runtime protection that is closer to flipping a switch than running an engineering project. Define your policy once, point the gateway at HiddenLayer, and every model behind it is covered.

To see how it works end to end, read the HiddenLayer integration guide in the TrueFoundry docs, or reach out to our team for a walkthrough.