Why it matters
Without AI screening, consignments risk misdescriptions, illicit origins, AML breaches, export license lapses, and seizure of lots due to non-compliant provenance and missing documentation.
From design to safety
Design
AI screening intent and policy
Define data standards for antiquities, ICOM lists, 1970 cut-off logic, and export history. Specify decision rules, thresholds, and audit requirements for compliant consignments; outline escalation paths and sign-off requirements.
- antiquities catalog
- ICOM Red Lists
- export history registry
- documentation schema
Build
Data ingestion and model setup
Ingest antique records, provenance notes, red list checks, 1970 cut-off indicators, and license data. Validate data quality, normalize fields, and establish risk thresholds with explicit exception paths and retry logic.
- consignment intake form
- provenance notes
- red list dataset
- export licenses
Operation
Live screening and flagging
Run continuous screening during intake and pre-sale; generate flags for provenance concerns, looted indicators, or missing export documents, triggering manual review and mandatory cite-and-archive notes.
- antecedent records
- condition reports
- lot photographs
Maintenance
Ongoing quality and compliance checks
Monitor data freshness, update Red List changes, track provenance updates, and refresh escalation workflows to maintain accuracy across consignments, including periodic revalidation after license renewals and item transfers.
- policy repository
- compliance calendar
- red list updates
Safety
Audit trails and regulatory handoffs
Log all AI decisions with user overrides; prepare export and AML documentation for regulatory inspection and internal governance reviews, including chain-of-custody and verifiable signer attestations.
- audit logs
- export documentation
- AML due diligence
Part by part
- 1
antiquities catalog
- Watch for
- Unlisted artifact in a contested period or culture with ambiguous provenance or hidden looted indicators that emerge from new research.
- AI action
- Flag for curator review and corresponding provenance documentation check with cross-reference to archival sources.
- 2
ICOM Red Lists
- Watch for
- Presence of a red-flag category: looted or illicit origin or recent addition to a list.
- AI action
- Run Red List cross-check and trigger escalation workflow with submission to provenance committee.
- 3
export history registry
- Watch for
- Export license missing or expired for a listed item or inconsistent export history trail.
- AI action
- Flag for compliance officer audit and license renewal, and log cross-border movement window.
- 4
documentation schema
- Watch for
- Incomplete or inconsistent provenance documents across records or conflicting attestations.
- AI action
- Prompt cataloguer to request missing documents and attach verifier notes with source citations.
- 5
conditions notes
- Watch for
- Discrepancies between condition report and provenance claims or signs of later restoration concealment.
- AI action
- Flag for independent expert re-valuation and reconciliation notes before sale.
- 6
lot photographs
- Watch for
- Missing or poor-quality images hindering provenance assessment or misalignment with catalog data.
- AI action
- Trigger required photography update, retake, and metadata tagging with high-resolution references.
Worked example
Audits of regional antiquities consignments
Illustrative scenario with sample data.
An auction house screens incoming consignments for provenance risk, ensuring compliance with ICOM lists and export controls before cataloguing.
Signal in
field: antiquity_id=Q-492, era=Hellenistic, looted_risk=high, red_list_status=blocked, export_license=none, 1970_cutoff=yes, provenance_source=unknown, previous_owner=unknown, catalog_entry_id=CE-310
System out
alert: high-risk antiquity Q-492; action: manual provenance verification, revoke cataloging until documentation is provided, escalate to provenance committee and archive interim hold.
Human sign-off: Head of Compliance signs off on the escalation and initiates a documented review with provenance team, with notes filed in the compliance diary and chain-of-custody record.
How to roll it out
Phase 1
Weeks 1 to 2
Baseline screening rules defined; data sources connected; initial auto-flags tested against live consignments with manual checks on a subset.
Phase 2
Weeks 3 to 6
Model tuning and escalation workflows integrated with catalogue platform and export licensing system; red list feeds synchronized with national heritage databases.
Phase 3
Weeks 7 to 12
Full auditability enabled; staff training completed; live-running with periodic compliance reviews, and quarterly governance audits documented.
Where people stay in charge
- Decisions about red-list status, looted indicators, or export risk must be validated by a qualified compliance officer with sign-off authority over provenance cases.
- All provenance determinations and export documentation must be auditable with immutable records retained for regulatory reviews and potential court proceedings.
- Manual overrides must be logged with justification and subject to periodic supervisory review and cross-departmental validation.
Questions owners and crews ask
How does AI screening affect provenance checks for consignments?
AI screening aggregates provenance notes, export history, and Red List status to surface risk flags; human investigators review flagged items and confirm admissibility before cataloguing, with traceable rationale and archival links.
Who signs off on red-list findings before sale?
The head of compliance or appointed compliance officer signs off on red-list findings after review of evidence and export licensing status, with a formal decision record.
What data sources are used for red-list screening?
The system reads ICOM Red Lists, export license databases, provenance records, and 1970 cut-off indicators to determine risk, cross-checking with origin-country registries and theft notices.
Can staff override automated decisions?
Yes, staff can override with justification; all overrides are recorded, reviewed by management, and subjected to subsequent audit as part of governance.