What this registry documents
A public, searchable registry of Danish content found in the open datasets used to train large AI models. It covers 24,792 Danish book titles found in the three shadow libraries documented in US litigation, 28,376 Danish music recordings, 25,938 song lyrics scraped from Genius, close to 5 million image references from Danish domains, 32 million Danish web pages in the training dataset FineWeb-2, and personal data belonging to ordinary Danish citizens, including 342 individuals whose names appear alongside their national ID numbers.
How the findings are verified
Danish ISBNs are matched against archived, contemporaneous copies of LibGen (June 2021), the Z-Library mirror (July 2022) and Books3, the exact versions that Anthropic was documented to have downloaded in Bartz v. Anthropic. Each of the 8,203 verified titles is anchored with the cryptographic MD5 hash of the file, so presence can be checked file by file. Web content is verified against FineWeb-2, an actual training dataset that anyone can inspect on Hugging Face. Every page of the registry states openly how strong the evidence is, and every count can be reproduced from the same public sources.
Why it matters outside Denmark
The 1.5 billion dollar settlement in Bartz v. Anthropic, granted final approval in July 2026, covers only works registered with the US Copyright Office. Danish works sat in the same files but fall outside the settlement. The same is true for most other non-American rights holders. Documentation of what was taken, from which dataset version, is the only durable evidence, and this registry shows what that documentation looks like for one small language.
About the registry
The registry is an independent project by Thor Martin Bærug, a music and media rights consultant based in Oslo. It is unaffiliated with any organisation and receives no funding. The registry itself is in Danish and lives at thormartinbaerug.com/ai-traeningsdata. Questions, corrections and press enquiries are welcome at thor@thormartinbaerug.com.