The DOI matches its Crossref record; 7 of the 7 references it cites were found in a registry or an index.
Towards Collecting Royalties for Copyrighted Data for Generative Models
Abstract
Addressing issues of copyrighted data in the context of generative models has become an important issue for content creators, publishers, organizations training generative models, and those who deploy generative models for particular applications. Copyright holders want to ensure that they are fairly compensated for their work and users of training data and models do not want to expose themselves to litigation. However, traditional models of bulk-licensing data fit only poorly the context of model training. In this paper, we want to discuss why a traditional data license is not always a good fit, how data is used in the life-cycle of generative models and which impact data has on model output. This can be used as a foundation for a pay-per-(model)use compensation based how data contributes to a model’s output. Having a way to compensate copyright holders in this way reduces risk for model trainers, avoids large investments upfront, and encourage a lively data ecosystem in which the creation and distribution of original work is encouraged and fairly compensated.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“Pre-training often uses multiple datasets, Common Crawl being only one of them, albeit typically the largest (Ludwig et al., 2024).”
From Generative AI and intellectual property: a systematic literature review of possible issues and mitigations (2026) · cites this paper as Ludwig et al., 2024
In the citing paper
“The datasets employed are much smaller and curated (Ludwig et al., 2024).”
From Generative AI and intellectual property: a systematic literature review of possible issues and mitigations (2026) · cites this paper as Ludwig et al., 2024
In the citing paper
“Organisations using crawled AI training datasets that are publicly available online typically do not pay compensation for these data sources, but assume that this is legitimate use under some legal doctrine in the jurisdiction in which the crawl takes place or the data source is located (Ludwig et al., 2024).”
From Generative AI and intellectual property: a systematic literature review of possible issues and mitigations (2026) · cites this paper as Ludwig et al., 2024
In the citing paper
“Data used at inference time, i.e., for Retrieval Augmented Generation (RAG) include data that changes frequently and cannot be meaningfully included in model training (Ludwig et al., 2024).”
From Generative AI and intellectual property: a systematic literature review of possible issues and mitigations (2026) · cites this paper as Ludwig et al., 2024
Citations
Cites
LeCun, Y., Bengio, Y., & Hinton, G. (2015) · Nature · doi:10.1038/nature14539 · Cited by 84,126
Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks
Lewis, P., Perez, E., Piktus, A., Petroni, F., Karpukhin, V., Goyal, N., Küttler, H., Lewis, M., Yih, W.-T., Rocktäschel, T., et al. (2020) · arXiv · doi:10.48550/arxiv.2005.11401 · Cited by 3,101
A Survey of Large Language Models
Zhao, W. X., Zhou, K., Li, J., Tang, T., Dong, Z., Hou, Y., Zhang, B., Min, Y., Zhang, J., Liu, P., et al. (2026) · Frontiers of Computer Science · doi:10.1007/s11704-026-60308-3 · Cited by 2,093 · cited 1×
“Large Language Models (LLMs) [2], [3] in particular have attracted much attention over the past years with the availability of ChatGPT to the general public, even if predictive language models were considered since the early days of computer science [4].”
BackgroundExplainability for Large Language Models: A Survey
Zhao, H., Chen, H., Yang, F., Liu, N., Deng, H., Cai, H., Wang, S., Yin, D., & Du, M. (2024) · ACM Transactions on Intelligent Systems and Technology · doi:10.1145/3639372 · Cited by 681
Large Language Models: A Survey
Minaee, S., Mikolov, T., Nikzad, N., Chenaghlu, M., Socher, R., Amatriain, X., & Gao, J. (2024) · arXiv · doi:10.48550/arxiv.2402.06196 · Cited by 325
Cross-Task Generalization via Natural Language Crowdsourcing Instructions
Mishra, S., Khashabi, D., Baral, C., & Hajishirzi, H. (2022) · Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · doi:10.18653/v1/2022.acl-long.244 · Cited by 269
Dolma: an open corpus of three trillion tokens for language model pretraining research
Soldaini, L., Kinney, R., Bhagia, A., Schwenk, D., Atkinson, D., Authur, R., Bogin, B., Chandu, K., Dumas, J., Elazar, Y., et al. (2024) · Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · doi:10.18653/v1/2024.acl-long.840 · Cited by 56
Research works printed but not linked: 5
- Rumelhart, 1986, Learning internal representations by error propagation, parallel distributed processing, explorations in the microstructure of cognition, ed. de rumelhart and j. mcclelland. vol. 1. 1986, Biometrika, 71, 599
- 10.1002/j.1538-7305.1951.tb01366.x · DOI · Paper page
- Rosati, 2018, The exception for text and data mining (tdm) in the proposed directive on copyright in the digital single market - technical aspects
- 10.1145/3626772.3657775 · DOI
- Marone, 2024, Data portraits: Recording foundation model training data, Advances in Neural Information Processing Systems, 36
Cited by
Arnaudo, A., Coppola, R., Morisio, M., Vetrò, A., Borghi, M., Raso, R., & Khan, B. (2026) · PeerJ Computer Science · doi:10.7717/peerj-cs.3928 · Cited by 2
Metadata
- Authors
- Ludwig, H., Zhou, Y., Zawad, S., Ong, Y., Li, P., Butler, E., & Zahid, E.
- Affiliations
- IBM Research,San Jose,CA,United States
- IBM Research,Cambridge,MA,United States
- Venue
- 2024 IEEE International Conference on Web Services (ICWS)
- Publisher
- IEEE
- Published
- 2024-07-07
- Pages
- 20-26
- Cited by
- 4
- References
- 7
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.
