Shadow AI and PHI Exposure · OnStak Use Case

Find the AI your hospital never approved.

Clinicians and staff reach for public AI tools outside any governance, and PHI can leak to a model with no BAA behind it. Nobody has a live view of which devices are calling which AI services, or what leaves the building. Here is how OnStak gave one health system that view, with evidence to show for it.

4-6 wk

POC, telemetry stays in your network

DPI

PHI and PII exposure assessment

HIPAA

evidence artifact, the Yes Layer

Security analyst reviewing a dashboard of unsanctioned AI traffic on a hospital network

Story highlights

  • A live view of unsanctioned AI use across the whole hospital estate.
  • PHI exposure, assessed and documented. Framed as HIPAA evidence from day one.
  • Built on the AI Correlation Fabric, correlating the telemetry you already have.

The situation

AI tools nobody approved, and nobody can see.

Clinicians and staff reach for public AI tools outside any governance, and PHI can leak to a model with no BAA behind it. Nobody has a live view of which devices are calling which AI services, or what leaves the building. They set out to see it before the regulator asks.

The problem

Exposure with no visibility and no evidence.

What you can't see, you can't govern. Three things drove it.

Shadow AI

Staff use public AI tools outside governance, and no one has a full view of it.

PHI can leak

Patient data sent to a model with no BAA is a HIPAA exposure.

No evidence

When the regulator asks, there is no record of what was allowed or what happened.

What OnStak did

Turn the telemetry you already have into a live AI inventory.

OnStak correlated existing telemetry, assessed exposure, and left behind evidence that holds up.

1

Deployed the AI Correlation Fabric

Correlate the firewall, network-monitoring and network-controller telemetry you already have.

2

Monitored for Shadow AI

Found the devices calling public AI services across the estate.

3

Assessed PHI and PII exposure

Using deep packet inspection to see what was actually leaving.

4

Designed enforcement and left evidence

Mapped to the controls you already run, with a HIPAA-framed evidence artifact behind it, the Yes Layer.

What changed

From invisible exposure to audit-grade evidence.

Before

Nobody could see what AI was in use.

  • Unsanctioned AI use no one can see.
  • PHI potentially leaving for public models.
  • No evidence for the compliance team or the regulator.
After, with OnStak

Now every device calling AI is visible.

  • A live view of every device calling an AI service.
  • PHI exposure surfaced and assessed.
  • Audit-grade evidence of what was allowed and what happened.

The outcome

Anonymized health system · Shadow AI and PHI exposure

Visibility, assessment and evidence, inside your own network.

4-6 wk
Proof of concept
Telemetry stays in your network.
DPI
PHI & PII exposure assessment
Via deep packet inspection.
HIPAA
Evidence artifact, the Yes Layer
Audit-ready documentation.
Self-hosted
Delivered on your own Kubernetes
All telemetry stays inside your environment.

The idea that holds it together

You make the call. We make it real.

“We start with your problem, and shape the solution around it.”

Awais Janjua  /  Chief Technology Officer, OnStak

Built on
AI Correlation Fabric (patent-pending) AIOps integration Deep packet inspection HIPAA evidence, the Yes Layer

Coffee? No Slides

No pitch deck. No 47-page proposal. Just a straight talk about what’s broken and what to fix first.

  • Solutions
  • Debrief
  • About Us