The intSignal AI Gateway is a private layer that sits between your applications and any AI model. Every prompt, document, and piece of context passes through it on the way to the model — and the model's response passes back through it — so sensitive data is detected and removed going out, and unsafe content is caught coming back in. The result: your teams and software can use AI on real business data without that data leaking to a third-party model.
It runs as its own internal AI/ML layer: purpose-built classifiers identify and sanitize data so you never have to hand raw customer records to an external model to get useful answers — the whole point of a compliance-safe AI gateway for business data rather than a policy you hope people follow.
What it does
- Redacts sensitive data — names, SSNs, card numbers (PAN/PCI), health details (PHI), emails, phone numbers, secrets, and API keys are detected and redacted, masked, or blocked before they leave. See Policies.
- Enforces your policies — the Gateway applies your organization's privacy and security rules to every request, including Microsoft Purview sensitivity labels, and blocks content that must never be shared (like password-reset and MFA-code emails).
- Works with any model — route to OpenAI (GPT), Anthropic (Claude), Azure OpenAI, or a private/on-prem model, and switch without changing your app. See Models & agents.
- Captures documents — scan-to-text and scan-to-form turn paper into structured, redacted data filed straight into your CRM, so intake goes paperless.
- Logs everything — a full prompt/response audit trail proves what was sent, what was redacted, and that customer data was never used for training.
Who it's for
| Audience | How they use it |
|---|---|
| Developers | Embed the Gateway with an SDK or REST API to sanitize prompts and responses inside their own product. |
| Integrators | Deploy it as a managed, multi-tenant proxy in front of the AI they build for clients. |
| Companies | Put it in front of the assistants and copilots their teams already use (ChatGPT, Microsoft Copilot). |
One protection, three ways to adopt it
The detection, policy, and logging engine is the same whether you call it as an SDK, run it as a reverse proxy, or point an existing assistant at it. Start with whichever fits your environment and change later without re-doing the work.
How it fits together
| Layer | Role |
|---|---|
| Your app / assistant | Sends prompts, documents, and context. |
| AI Gateway | Detects and redacts sensitive data, enforces policy, logs the request, then forwards a sanitized prompt. |
| AI model | GPT, Claude, Azure OpenAI, or a private model — receives only sanitized input. |
| Response check | The model's reply passes back through the Gateway before reaching your app. |
Next steps
- Stand up your first sanitized call in Getting started.
- Understand the flow in How it works.
- Define what to redact and block in Policies and let the Gateway propose rules with Automatic policy setting.
- Connect models and integrations, and review Privacy & compliance.
- Automate everything with the AI Gateway API.