Cyber SecurityCloud
1-888-984-1634
Customer loginPartner login

Product · AI governance & privacy

intSignal AI Gateway

Put AI to work on real business data — without leaking any of it.

AI is only safe for business when the data it sees is safe to share. The intSignal AI Gateway is a private layer that sits between your applications and any AI model: it detects and redacts sensitive data, enforces your organization's privacy and security policies, and blocks anything that shouldn't leave — before a single token reaches a model. Software teams embed it, integrators deploy it for clients, and companies run it in front of the AI tools they already use.

Request a demo  ⟶← All products
Live, interactive demo — click around to explore.
Redacts PII · PHI · PCIBlocks sensitive emailEnforces your policiesWorks with any modelScan-to-text & formsFull audit log
Customer email · inbound
From jordan.rivera@acme.com· 9:41 AM
Re: refund on my last order
Hi — here are my details so you can process the refund:
Namename•••••• ••••••
SSNssn•••-••-••••
Cardpan · pci•••• •••• •••• ••••
DOBdob•• / •• / ••••
Phonephone(•••) •••-••••
Note Password-reset link — blocked, never sent to AI
intSignal AI Gateway
5 entities redacted · 1 blocked · safe for AISanitized

How it works

A sanitizing layer that sits between your work and the AI

The gateway is an inline broker. Prompts, documents, and context pass through it on the way to the model, and the model's response passes back through it — so sensitive data is stripped going out and nothing unsafe comes back in. It's an internal AI/ML layer of its own: purpose-built models classify and clean the data, so your business data never has to be handed over raw to a third-party model to be useful.

✓Inline by designDetect, redact, and enforce policy on every request — not after the fact.
✓Any modelOpenAI, Anthropic, Azure OpenAI, or a private/on-prem model — swap freely.
✓Two-wayOutbound prompts are sanitized and inbound responses are inspected.
✓Deterministic + AIPattern detectors plus ML classifiers catch what regex alone misses.
Where the gateway sits
Your app
Copilot / chatbot
Agent workflow
intSignal AI Gateway
detect · redact · enforce · log
OpenAI · Anthropic
Azure OpenAI
Private / on-prem LLM
inlineNothing reaches a model until it’s sanitized — and the reply is checked on the way back.

Redaction

Finds and removes sensitive data before it ever leaves

Customer records, SSNs, card numbers, health details, credentials, and API keys are detected and handled by type — redacted, masked, or blocked — with the original values kept safely on your side and restored in the response when it's appropriate.

✓Regulated dataPII, PHI (HIPAA), and cardholder data (PCI/PAN) recognized out of the box.
✓Secrets & keysPasswords, tokens, and API keys blocked so they never reach a model.
✓Reversible tokensRedacted values are re-linked in the reply when the workflow needs them.
✓Tunable per fieldChoose redact, mask, or block for each data type and use case.
Detected & handled
SSN / TINRedact99%
Card number (PAN) · PCIMask99%
Health info · PHIRedact97%
Email & phoneMask98%
Names & addressesRedact96%
Secrets & API keysBlock99%
detectTuned detectors for regulated data — redact, mask, or block per type.

Policy

Enforces your applications' security and privacy policies

The gateway applies the same rules your systems already live by. It honors platform data-protection controls — including Microsoft's sensitivity labels and redaction capabilities — and adds guardrails of its own: it won't read a password-reset or MFA-code email, won't send cardholder data to a model, and never lets customer data be used for training.

✓App-awareRespects Microsoft Purview / 365 sensitivity labels and redaction.
✓Content guardrailsBlocks password-reset, MFA codes, and other must-not-share messages.
✓Compliance-mappedPolicies aligned to HIPAA, PCI DSS, GDPR, and SOC 2 controls.
✓No training on your dataPrompts and content are never used to train third-party models.
Active policy · Acme default
Honor Microsoft Purview sensitivity labels✓ Enforced
Block password-reset & MFA-code emails✓ Enforced
Mask PAN before any model call (PCI DSS)✓ Enforced
Redact PHI on non-clinical prompts (HIPAA)✓ Enforced
Never use customer data to train models✓ Enforced
Full prompt / response audit trail✓ On
enforceYour organization’s privacy rules, applied to every prompt — the same way the app would.

Documents

Scan to text, scan to form — go paperless

Point it at a scan, a PDF, or a photo and it turns paper into structured data: it reads the document, extracts the fields, redacts what's sensitive, and files the result straight into your CRM or line-of-business app — so intake that used to be manual becomes automatic and paperless.

✓Scan-to-text (OCR)Accurate text extraction from scans, PDFs, and photos.
✓Scan-to-formMaps extracted fields into CRM records and forms automatically.
✓Redacted on captureSensitive fields are masked as documents are ingested.
✓Paperless intakeRoute the structured result to your CRM, ticketing, or ERP.
Scan → extract → CRM
CompanyAcme Manufacturing
DocumentSigned MSA
Amount$14,200 / yr
EffectiveMay 1, 2026
✓ Filed to CRM · paperless
captureScan-to-text and scan-to-form — extracted, redacted, and filed automatically.

For builders

Built for developers, integrators, and teams

Ship AI features without shipping a data-leak risk. Developers embed the gateway with an SDK or REST API; integrators deploy it as a managed proxy in front of their clients' AI; and companies drop it in front of the assistants and copilots they already use — same protection, three ways to adopt it.

✓For developersSDK + REST API to sanitize prompts and responses inside your product.
✓For integratorsDeploy a managed, multi-tenant proxy for the clients you build for.
✓For companiesPut it in front of ChatGPT, Copilot, or any assistant your teams use.
gateway.ts
// one call — sanitized, policy-checked, and logged
const safe = await gateway.sanitize(prompt, {
  policy: "acme-default",
  redact: ["pii", "pan", "phi", "secrets"],
});
const reply = await llm.complete(safe.prompt);
// → 5 entities redacted · 0 leaked
SDKDrop-in SDK & REST API — or a managed proxy in front of any model.

Full platform

Everything in the AI Gateway

Detection, policy, delivery, and document capture — grouped. Expand any area for the full list.

Detection & redaction10
PII (names, addresses, SSN/TIN)PHI (HIPAA)Cardholder data / PAN (PCI)Emails, phones, DOBSecrets, tokens & API keysCustom entity patternsML + pattern detectorsRedact / mask / block per typeReversible tokenizationResponse inspection
Policy & compliance8
Microsoft Purview sensitivity labelsContent guardrails (password-reset, MFA)HIPAA · PCI DSS · GDPR · SOC 2 mappingNo-training guaranteePer-app & per-team policiesData residency controlsHuman-in-the-loop reviewPrompt & response audit trail
Models & delivery8
OpenAI & AnthropicAzure OpenAIPrivate / on-prem modelsSDK (drop-in)REST APIManaged reverse proxyMulti-tenant for integratorsLatency-optimized inline path
Documents & capture8
Scan-to-text (OCR)Scan-to-form field extractionPDF & photo intakeRedaction on captureCRM / ERP / ticketing filingPaperless workflowsTemplate mappingBulk ingestion
Operations & governance8
Full audit loggingRedaction reportsAlerting on leaks & policy hitsRole-based accessSIEM exportUsage & cost analyticsSSO (Entra ID)On-prem or cloud deployment

See your data get sanitized in real time

Bring a real prompt or document and we'll show you exactly what the gateway detects, redacts, and blocks — before it reaches a model.

Request a demo  ⟶

Get started

Request a demo of AI Gateway

Tell us a little about your environment and a specialist will walk you through AI Gateway on a live demo — and put together pricing built for your scale.

Prefer to talk now? Call 1-888-984-1634.

Frequently asked questions

What does the AI Gateway actually do?

It sits between your applications and any AI model as an inline broker. On the way out it detects and redacts sensitive data — PII, PHI, cardholder data (PCI/PAN), and secrets — and enforces your privacy and security policies; on the way back it inspects the model's response. The result is that your teams and software can use AI on real business data without that data leaking to a third-party model.

Does it work with the AI models and tools we already use?

Yes. It's model-agnostic — OpenAI, Anthropic, Azure OpenAI, or a private/on-prem model — and you can put it in front of assistants and copilots your teams already use, embed it in your own software via SDK/REST API, or run it as a managed proxy.

How is this different from just turning on a model's privacy setting?

Model settings can't reliably find and remove sensitive data in your prompts, enforce your organization's policies, or block content like password-reset emails. The gateway is a dedicated internal AI/ML layer that classifies and sanitizes data before it's ever sent, and applies the same rules your apps live by — including Microsoft Purview sensitivity labels and redaction.

Who is it for?

Three audiences: software companies that want to build privacy-safe AI into their products, integrators that deploy AI for their clients, and regular businesses that want their teams to use AI assistants safely. Same protection, delivered as an SDK, a managed proxy, or an in-front-of-your-tools deployment.

What are the scan-to-text and scan-to-form features?

The gateway can turn paper into structured data: it reads a scan, PDF, or photo, extracts the fields, redacts anything sensitive, and files the result into your CRM or line-of-business app — so document intake becomes automatic and paperless.

Is our data used to train AI models?

No. Prompts, documents, and content processed by the gateway are never used to train third-party models, and everything is logged so you can prove it at audit time.

Related products & services

intSignal Compliance PlatformintSignal Zero TrustData loss prevention (DLP)Machine Learning & AIWorkforce Modernization