Loom AI Labs

live Privacy and security

Privacy Shield

A local-first support prototype that prepares privacy-aware context before retrieval or generation.

Privacy Shield live demo interface
Live application capture · verified 2026-07-25
Status
live
Last verified
2026-07-25
AI Fabric
0.4.0
Data
Synthetic

Question

Hypothesis

Can privacy preparation become a visible application stage rather than an invisible prompt instruction?

What this demonstrates

  • Input and output detection direction is configurable.
  • Redaction and audit behavior can be inspected without an OpenAI key.

What this does not demonstrate

  • Legal compliance
  • Coverage of every identifier or jurisdiction

Architecture and data flow

  1. 01 Synthetic source data
  2. 02 Host application policy
  3. 03 AI Fabric capability
  4. 04 Observable UI result

Guided steps

  1. 01

    Submit the provided synthetic support message.

  2. 02

    Compare detected values with the prepared context.

  3. 03

    Inspect the local audit result.

Observable evidence

The running application and its health surface are the public evidence for this experiment. The source repository exposes the implementation path behind the visible result.

Security and data boundary

The application selects detection mode, encryption posture and downstream policy. The demo intentionally does not require a remote model.

The sample personal identifiers are synthetic and disposable.

Known limitations

  • Pattern detection can miss novel identifiers.
  • The demo is not a legal or compliance assessment.

Technologies and products used

  • Spring Boot
  • AI Fabric 0.4.0
  • Local detection
  • Local embeddings

Related research

The investigation behind this proof.

Privacy and identity Implemented prototype

Privacy-Aware Context Preparation for RAG

This investigation treats privacy preparation as an application-visible stage with configurable direction, storage posture and audit evidence.

Artifact
Security investigation
Maturity
Implemented prototype
Read investigation

Connect

Inspect the live system and its source together.

The experiment is useful when the visible result, application boundary and implementation can all be questioned.