← AI Data Privacy & PII Management

The Gateway Pattern: Sanitise Before Sending

The architecture that unblocks enterprise AI adoption — local PII detection, redaction, cloud AI processing, and re-hydration of pseudonymised tokens back to originals.

The single most impactful pattern

If you take one architecture from this entire course, this is the one. The gateway pattern is the reason enterprises that were blocked on AI adoption for years are now deploying AI across their organisations. It resolves the fundamental tension — AI needs data, privacy demands restriction — by processing the data locally before it ever leaves your environment.

The architecture is straightforward:

Raw Data → Local PII Detection → Redaction/Pseudonymisation → Cloud AI → Re-hydration → Response

Your user types a prompt containing sensitive data. Before that prompt reaches any cloud AI service, a local gateway intercepts it, detects PII using the pipeline from Module 4, redacts or pseudonymises the PII using the techniques from Module 5, and then forwards the sanitised prompt to the cloud AI. When the response comes back, the gateway replaces pseudonymised tokens with the originals and delivers the complete response to the user.

The cloud AI never sees the sensitive data. Your audit log captures what was detected and how it was handled. Your users get the full power of cloud AI capabilities. Your compliance team gets the privacy guarantees they require.

?

An enterprise has been blocking AI adoption for 18 months due to data privacy concerns. The CISO wants zero sensitive data leaving the environment. The CEO wants AI-powered productivity gains. Who does the gateway pattern satisfy?