When War Zones Empty Museums,
British Museum • Antiquities Tracking System
Where Do the Treasures Go?
In war-torn regions like Palmyra, historic antiquities don't just disappear, they surface on auction websites and vanish forever into private collections, in a lucrative illicit trade. For the British Museum, the challenge wasn't just knowing where this was happening and what was being sold, it was to create a solution that could provide the evidence that could stand up to scrutiny in prosecution.
Adappt was commissioned to build an AI-powered system that monitored global auction platforms, identified looted artifacts, tracking the networks selling them, and collected evidence before listings disappeared.
The result: 20 terabytes of data revealing the scale of the organised crime behind antiquities trafficking.
This is what we learned about protecting cultural heritage against organised criminal activities.
The Evidence Disappears Faster Than the Artifacts
War creates opportunities for organised crime. When countries destabilise, such as Syria, Iraq, Libya, archaeological sites become targets. Gangs loot museums and dig sites, taking ancient Mesopotamian figurines, Greek pottery, Egyptian artifacts. Items with both significant cultural and monetary value.
These stolen antiquities often surface on antiquities auction websites, world-famous auction houses and sometimes platforms as open as eBay or Facebook Marketplace. They're listed, auctioned, sold within very short periods of time, and then the listings disappear. The evidence will have gone, and the artifacts will have been lost forever to private collections.
The Challenge:
Traditional methods couldn't keep pace. By the time researchers manually identified a suspicious listing, it had already sold and disappeared. They needed to capture evidence as it appeared, before it vanished.
But the scale was overwhelming:
- Hundreds of auction platforms worldwide
- Listings posted and removed within hours or days
- Massive volumes of traffic and content
- Sellers changing names and accounts regularly
- Items described in ways that obscure their true origin
- Photographs that might, or might not, match museum databases
And critically, the team needed evidence that would support further action where illegal activity had been identifed. This included: Screenshots. Seller information. Account connections. Timestamps. A complete chain of documentation.
Pattern Recognition at Scale
We built a specialised scraping engine designed to understand and recognise antiquities across global auction platforms. This highly sophisticated engine included layered entity recognition and rule-based pattern matching to identify stolen and looted artifacts among millions of listings.
The Technical Approach:
At its core, the system needed to solve several problems simultaneously:
Entity Recognition - Identify antiquities among vast amounts of auction content
Pattern Matching - Recognise descriptions that indicate looted origin (specific regions, time periods, provenance gaps)
Evidence Capture - Take screenshots and collect complete listing data before it disappears
Seller Tracking - Connect accounts across platforms even when sellers change names
Network Mapping - Build relationships showing how items and sellers connect
How It Worked:
The system continuously monitored auction websites, from major international houses to smaller platforms, look for specific item classifications, terms and providence. When a potential match was identified, the platform captured all available evidence, including: listing text, photographs, seller information, pricing, timing.
This provided the key raw data needed for the analyst team, but the real innovation was in connecting the dots clues that they left behind. Sellers who traffic stolen antiquities change their account names regularly, but by analysing patterns, listing styles, timing, geographical indicators, payment methods etc. the system was designed to connect accounts and say with confidence: "This is the same person."
Over time, the system would reveal networks. We already knew that items which appeared in Amsterdam, could swiftly move to New York, and may end up in Johannesburg. The system's goal was to help identify the distribution routes where stolen artifacts that were being used by organised crime.
Twenty Terabytes of Evidence
During its commissioning, the platform continuously gathering a vast amount of evidence, revealing the scale of the problem:
20 terabytes of data
The platform conducted systematic evidence building, with each listing including:
The Vision: A Global Transit Map
The British Museum's wider ambition, had funding continued, was to create a global map to visualise transit lines and timelines across a globe showing where antiquities originated, which collections they moved through, which cities they passed through and where they ultimately ended up. In short, to build a comprehensive view of organised antiquities crime.
The Impact on Their Team:
The researchers working on the project reported that the platform exceeded expectations. Before the system was introduced, a small team was resourced to manually monitor these auction platforms. The scale of the task was vast. After deployment, the team could readily identify which items were worth chasing and which weren't. The intuitive dashboard, included workflow tools, sharing and assigning options and evidence curation. The time taken to review potential illicit artifacts was dramatically reduced, whilst their operational effectiveness increased significantly.
This Isn't Opportunistic Crime
When we started, we imagined thieves grabbing artifacts and selling them online was opportunistic. The reality is far more sophisticated. This is organised crime operated by gang networks, with systematic looting coordinated across borders and mature supply chains that move stolen artifacts through specific routes and specific auction houses.
War zones don't just create opportunities for random theft, they create room for established criminal networks that know exactly how to move antiquities from rubble to private collections.
The Project Ended
The project funding eventually ended, with Adappt transferring some 20 terabytes of evidence. to the British Museum. The system proved what's possible in the world of comprehensive evidence collection at scale, and in revealing networks that were previously invisible.
""I am always impressed with the professionalism, expertise, imagination and sheer enthusiasm of the people at Adappt, which has made it such a pleasure to work with them on a project to promote cultural heritage protection."
A Thank You
When the project concluded, the British Museum's main Egyptologist took Adappt on a private tour of the vaults beneath the museum, to the collections not on public display.
We were privileged to be with someone who was completely fluent in ancient hieroglyphs. We were shown a hieroglyph that worked like an ancient crossword puzzle, that could be read in three different ways, each revealed different meanings. Even the direction figures faced changed the interpretation.
It was a glimpse into the depth of knowledge required to understand what the institution is trying to protect. These aren't just "artifacts", they're pieces of human history, storytelling, culture, knowledge systems that took centuries to develop and can be lost in moments.
Building for Evidence, Not Just Detection
Why Pattern Recognition Over Computer Vision:
You might expect an antiquity tracking system to rely heavily on computer vision, matching photographs of stolen items to museum databases. While image recognition played a role, the real challenge was identifying listings in the first place among millions of auction postings.
That required sophisticated pattern recognition. Certain phrases, geographical references, time periods, provenance descriptions (or suspicious lack thereof), seller behaviours, layered together to flag potential stolen items.
The Scraping Challenge:
Auction websites aren't designed to be scraped at scale. They have anti-bot protections, rate limits, dynamic content loading and usage terms and conditions. Building an engine that could reliably monitor hundreds of platforms continuously, respecting all legal obligations under their terms of usage, required constant adaptations depending on the sites in question.
The Evidence Standard:
Academic research and casual monitoring have different standards than evidence that might inform law enforcement or international heritage repatriation efforts. Every piece of data needed timestamps, provenance, screenshot verification, and chain of custody documentation. The system was built from the start with evidentiary standards in mind.
The Scale of Antiquities Trafficking
Cultural heritage crime is estimated at billions of dollars annually, rivalling arms and drug trafficking in organised crime revenue. The Syrian conflict alone saw massive looting from sites like Palmyra, with ancient artifacts flooding international markets.
The challenge isn't just catching individual sellers, it's understanding and disrupting the networks that make systematic trafficking possible. That requires data at scale, pattern recognition across borders, and evidence that can support international cooperation.
Need to Track and Analyse Data at Scale?
The British Museum project demonstrates capabilities that apply beyond cultural heritage: large-scale data harvesting, pattern recognition across noisy datasets, connecting disparate accounts and entities, building network visualisations, collecting evidence that meets rigorous standards. Whether you're tracking financial fraud, analysing supply chains, monitoring compliance, or understanding complex networks, the same technical approaches apply.