About Consortium AI

Trade finance fraud is a cross-bank problem: the same forged invoice, shell counterparty, or duplicated bill of lading surfaces at multiple institutions — but each bank only sees its own slice. Consortium AI is a bank-node deployment that lets institutions pool fraud risk signals without pooling customer data.

How it works

Each bank runs its own node. Invoice metadata — extracted by an AI model or entered manually — is gated against the bank's own confidence threshold before anything happens. Every critical field must individually meet the threshold; a missing confidence score fails, and a high overall score can never mask a failing field. High-confidence transactions are anonymized and shared with the consortium automatically. Everything else waits for a registered human reviewer.

Identifying fields never leave the node in clear text — they are shared only as SHA-256 hashes, so consortium members can match signals (the same supplier appearing at three banks) without learning who the supplier is. Amount, currency, and date stay clear for cross-bank risk scoring.

Compliance first

The system is built to fail safe, never open. Review decisions require a registered, active reviewer and land in an append-only audit trail. Untrusted document text is treated strictly as data on its way into any AI model — extraction output is re-validated and re-gated exactly like manual input, so the model can never grant itself auto-approval.

Where it stands

Consortium AI runs live as a single bank-node with the full gating, review, anonymization, and audit pipeline. The multi-bank consortium transport is pluggable and currently points at a reference implementation. See the changelog for what has shipped, or get in touch.