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
- 01 Synthetic source data
- 02 Host application policy
- 03 AI Fabric capability
- 04 Observable UI result
Guided steps
- 01
Submit the provided synthetic support message.
- 02
Compare detected values with the prepared context.
- 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.