AI transparency

AI is a layer inside the evidence system. It does not replace the decision.

EmlakIQ outputs are anchored in a deterministic data pipeline. When an AI-assisted component is present, that fact is stated explicitly together with scope and confidence context.

Guardrail

  • AI does not make the decision; it makes evidence readable.
  • Out-of-scope cases trigger disclosure rather than synthetic output.
  • Confidence stays attached to the model layer.

Governance rules

The model layer only operates within these boundaries.

Evidence-bound explanation

Any AI-assisted explanation must show which fields it relies on. Unsourced interpretation is not acceptable system behavior.

Fact and interpretation split

Fields from sources or deterministic transformation are not blended with model commentary. The two layers stay visibly separate.

Scope disclosure

A model must state which geographies or data classes it applies to. It does not silently operate outside supported scope.

Limit disclosure

The system does not produce official valuation, legal opinion, or structural safety certification. Those boundaries are shown directly to the user.

Snapshot discipline

Summaries are stored with snapshot identity so the same data state can be inspected again with a traceable result.