Detect
Local classifiers identify personal, commercial and organisationally sensitive information.
Case study · Secure enterprise AI
Sensitive data is masked before the model sees it.
A local gateway anonymises prompts, applies company policy and records every action before using an external model.
Operating proof
3Classification layersApplied before content leaves the organisation
LocalSensitive-data detectionOriginal content stays in client infrastructure
EUModel processingWith a complete audit trail
02 / Problem
Blocking public AI did not remove the demand for it.
Employees still needed generative models for everyday work, creating the risk that customer data, commercial terms or other sensitive information would be pasted into external tools without organisational control.
03 / System
A controlled gateway between employees and external models.
The gateway runs inside client infrastructure. It detects and masks sensitive content, selects a model according to task, cost and policy, and keeps a complete record of the decision path.
Approve
The user sees the anonymised version and remains in control before anything is sent.
Route
The request is matched to an approved model under the organisation's policy.
Audit
Inputs, transformations, routing and responses are recorded for review.
04 / Outcome
Employees can use powerful models without exposing the original data.
The gateway was completed and transferred to BLIK. Sensitive information is anonymised before external processing, model work remains in the EU, and every action enters a complete audit log.