Stolen Art Databases and the Limits of a Clean Search

Three registers that are not doing the same job

The Art Loss Register was established in London in 1990, growing out of an art theft archive created by the International Foundation for Art Research in 1976. It is a private company registered in England and Wales under number 02455350, and it describes its database as the world's largest private database of lost, stolen and looted art, with more than 700,000 items and roughly 450,000 searches conducted each year. Entries are made on behalf of victims of theft and looting, insurers, police forces and others, and police, the trade and the public may all record items. Its services are free to law enforcement agencies and nation states; the auction houses and dealers who use it subscribe.

Interpol's Stolen Works of Art database is a different animal. It holds almost 57,000 items with descriptions and pictures, and entries may be submitted only by Interpol National Central Bureaus and specified international partners including UNESCO, ICOM and ICCROM. Submissions must concern fully identifiable objects. Anyone can apply to become an authorised user and search it in real time, including by image matching. It is police information, filtered through national police forces.

The Lost Art Database, operated by the German Lost Art Foundation, is not a theft register at all. It carries two kinds of entry. A search request is published by an individual or institution for cultural property taken through Nazi persecution or lost in the Second World War. A found-object report is published by a holder whose object is identified as Nazi-looted property, or whose provenance gaps suggest that it may be. In other words, one half of the database is people looking for things and the other half is holders declaring uncertainty about things they have. As of January 2022 it recorded 11,165 objects as restituted.

A theft register cannot record a forced sale

This is the structural point from which everything else follows. The paradigmatic Nazi-era loss is not a burglary. It is a sale executed under duress, often for a real price paid into a blocked account, often with paperwork, often in order to obtain an exit visa. The Cassirer Pissarro is the canonical instance: a compelled sale to a Berlin dealer in 1939 for 900 Reichsmarks, followed after the war by a settlement with the Federal Republic. No police force was ever asked to open a file. There is no crime number, no incident report and no insurer subrogated to the loss.

An object dispossessed that way cannot enter a register that is built out of reported thefts, because the qualifying event never occurred in the form the register recognises. That is not a defect in the Art Loss Register or in Interpol. It is a boundary condition of what those institutions are.

The German Lost Art Database exists because of that boundary. Its found-object report is a museum publishing a gap in its own knowledge, which is an act with no analogue in policing. A curator who reports an object because its provenance for 1933 to 1945 is unclear is not alleging a crime and not naming a suspect. That is why the entry is made in the language of research rather than of law enforcement, and why the German Lost Art Foundation directs users to a separate legal FAQ rather than making any claim about what an entry establishes.

Registration is voluntary, and the gaps are systematic

Coverage in every one of these systems tracks who had the standing, the knowledge and the incentive to file. Interpol takes only what a national police force forwards through its National Central Bureau, and only for objects that are fully identifiable, which excludes exactly the undocumented and unphotographed material that dominates historical losses. The Art Loss Register takes what victims, insurers and the trade choose to register, so its density follows insurance penetration and market value. A family dispossessed in Vienna in 1938 had no insurer to make a claim, no police force that would take the report, and in many cases no photograph of the object.

The research corpora fill part of the gap and have their own limits. The database of the Einsatzstab Reichsleiter Rosenberg, jointly administered by the Claims Conference and the United States Holocaust Memorial Museum, covers more than 40,000 objects taken from Jews in German-occupied France and Belgium, reconstructed from surviving registration cards and photographs. The Central Registry of Information on Looted Cultural Property 1933-1945 lists over 25,000 objects from fifteen countries. The American Alliance of Museums' Nazi-Era Provenance Internet Portal listed some 30,000 objects across a hundred and seventy-nine museums, and operated from 2003 until 2024; it is no longer actively maintained. The Legacy Explorer database launched in December 2025 with the aim of digitally uniting archival evidence of Nazi-era theft that remains scattered across archives worldwide, which is a candid statement of the current position.

These are not a federated system. They do not share identifiers, they do not cross-query, and each covers a different slice of geography, period and object type. Searching one of them and reporting the result as a provenance check is a category error.

What a clean certificate says, and what it is worth

A negative search result is a statement with a precise and narrow meaning: on the date of the search, using the identifiers supplied, the object was not matched against the records then held by that one organisation. It is not a warranty of title, not title insurance, and not a provenance. Its accuracy depends on the quality of the descriptors given to the searcher and on the quality of the original entry, which may have been made decades earlier from a poor photograph and a vague description.

Understood correctly, it still has real value, and the value is best stated in the vocabulary of the law of title rather than of security. A documented search is evidence of due diligence. Under article 4 of the UNIDROIT Convention it is part of what converts a possessor who must hand the object back into a possessor who must be compensated for handing it back. In the civil law systems it is part of what makes good faith provable, and good faith is the precondition of German ten-year prescription and of Spanish three-year prescription under article 1955. In a jurisdiction applying nemo dat it protects the buyer's conscience and the buyer's claim against the seller, and it does not protect the buyer's title.

What it never does is tell the buyer that the object is unencumbered. The categories of loss it cannot see are the categories that generate the largest claims.

The tribunal now hearing these claims starts from the gap

The clearest measure of what a clean search is worth is what the deciding body does with it. Since 1 December 2025 claims against public institutions holding Nazi-era objects in Germany are heard by the Court of Arbitration for Nazi-Looted Cultural Property, administered by the German Lost Art Foundation, with binding awards. Its assessment framework treats gaps in provenance as unavoidable in light of the passage of time and the circumstances of the Holocaust era. It accepts indirect evidence, circumstantial evidence, historical contextualisation and prima facie reasoning from typical patterns. It requires a high degree of probability rather than certainty. And it applies a presumption that a sale loss resulted from Nazi persecution, subject to limited exceptions for certain flight transactions.

Read that against the reflex of the market. A holder who says the object was searched and came back clean is offering the absence of a record to a tribunal that has expressly decided the absence of records is the normal condition of these cases and has built its evidentiary rules around it. The clean search neither rebuts the presumption nor fills the gap; it simply confirms that nobody filed a police report, which everyone already knew.

The practical instruction is to stop treating database searches as diligence and start treating them as the first ten minutes of it. The work that decides an object's status is the reconstruction of where it physically was, and in whose hands, across the years that matter, out of exhibition catalogues, dealer stock books, auction records, shipping and customs papers, restitution files and the surviving archives of the confiscating agencies. The registers tell you whether somebody once reported a loss. Only the archive tells you what happened.