Why it matters
Without AI screening, misdescribed or stolen works may proceed to sale, risking AML breaches, provenance gaps, and potential seizure or reputational harm for the house.
From design to safety
Design
AI feasibility in risk screening
Assess data sources (Stolen Art Register, Art Loss Register, Interpol), data schemas, and decision thresholds; define retention, audit, and escalation policies for flagging potential matches during sale planning.
- Stolen Art Register ledger
- Art Loss Register subscription
- Interpol SI database access
Build
Integrating data sources
Ingest and normalize lot images, descriptions, and certificates; configure multi-source match rules, anti-tamper checks, watermark verification, and audit logging for intake review and provenance ramp.
- Lot images archive
- Description field mappings
- Search certificates catalog
Operation
Real-time screening in workflow
AI runs rapid comparisons on incoming consignments, generates risk scores, cross-checks watchlists, and routes flagged items to the compliance review queue with timestamped logs.
- Lot images archive
- Description field mappings
- Interpol SI database access
Maintenance
Data quality and rule tuning
Regularly update records, verify certificate validity, retrain similarity models, refresh watchlists, and reconcile overrides; run monthly data quality reports to maintain accuracy across holdings.
- Stolen Art Register ledger
- Art Loss Register subscription
- Search certificates catalog
Safety
Provenance and export compliance checks
Ensure flagged items undergo AML due diligence, provenance validation, export controls, and seizure risk assessments before listing; require sign-off by Head of Compliance with documented escalation.
- Interpol SI database access
- Lot images archive
- Description field mappings
Part by part
- 1
Stolen Art Register ledger
- Watch for
- Flagged matches with high similarity scores or shifted provenance indicators during intake and cataloging.
- AI action
- Read: image and description similarity; Decide: escalate for compliance review if confidence exceeds threshold.
- 2
Art Loss Register subscription
- Watch for
- Presence or absence of listed items; confirm alias titles and historical owners.
- AI action
- Read: cross-reference with registered entries; Decide: mark as potential risk or clear if no hit.
- 3
Interpol SI database access
- Watch for
- Cross-border ownership flags or pending notices tied to consignments.
- AI action
- Read: Interpol alerts; Decide: route for provenance investigation.
- 4
Lot images archive
- Watch for
- Discrepancies between on-file images and described lot features.
- AI action
- Read: image hashes and feature descriptors; Decide: request new photography or re-photography if needed.
- 5
Description field mappings
- Watch for
- Inconsistent or missing provenance notes, dates, or maker attributions.
- AI action
- Read: text fields; Decide: suggest standardized terms and flag gaps.
- 6
Search certificates catalog
- Watch for
- Expired or invalid search certificates accompanying consignments.
- AI action
- Read: certificate IDs and expiry dates; Decide: trigger renewal or verification workflow.
Worked example
Screening a regional pottery collection for provenance risk
Illustrative scenario with sample data.
A regional auction house processes an incoming consignor with multiple lots; AI screens for theft links before catalogue creation.
Signal in
lot_id: 4726; item_title: '18thC Seated Bearer Vase'; images: 256x256; description: 'Delft style, maker unknown'; certificates: 'SC-2024-075'; source_country: 'UK'
System out
risk_score: 0.78; flags: ['Interpol match: potential notice','Certificate expiry: 2025-11-01']; action: 'compliance review queue'; notes: 'requires provenance documents from consignor.'
Human sign-off: Head of Compliance reviews flagged items, approves or assigns to provenance team for further verification.
How to roll it out
Phase 1
Weeks 1 to 4: Baseline
Establish data feeds, test matching against a small batch, calibrate thresholds, and set up audit logs.
Phase 2
Weeks 5 to 8: Expansion
Ingest full catalog of current consignments, refine rules, and integrate with consignor intake forms.
Phase 3
Weeks 9 to 12: Live monitoring
Operate in production with live consignments, generate risk reports, and enforce escalation protocols.
Where people stay in charge
- Only a qualified compliance officer signs off on provenance validity and AML flags before listing.
- All high-risk matches must trigger a full provenance investigation with documented escalation and supervisor approval.
- Export controls and disposition decisions require sign-off by the Head of Compliance or managing director.
Questions owners and crews ask
What records does the AI screen compare against when screening consignments?
It compares lot images, descriptions, and attached search certificates against stolen art registers, Art Loss Register, and Interpol databases, producing a risk score and flags for review.
Who approves a flagged lot before it can be listed?
The Head of Compliance or designated compliance officer must review and approve the provenance and AML checks before listing.
How are certificates and certificates validity handled?
The system checks certificate IDs and expiry dates, and prompts renewal or verification if certificate data is invalid or expired.
Can the system trigger trace investigations or export controls?
Yes, for flagged provenance or AML concerns, it generates an escalation to the compliance team and records the decisive action.