Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat
Sign Up
Cited by 6

The DOI matches its Crossref record; 6 of the 6 references it cites were found in a registry or an index.

Tackling copyright issues in AI image generation through originality estimation and genericization

Chiba-Okabe, Hiroaki, Su, Weijie J.·2025·Scientific Reports
DOI 10.1038/s41598-025-90827-1PMC PMC11950165
Open this paper in CoChat — free account

Overview

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1038/s41598-025-90827-1

Cites

6 of 6 references verified0 citation errors · 0 not checkable

Statistics

Figures

Algorithm 1 PREGen.
Algorithm 1 PREGen.CC BY-NC-ND · PubMed Central
Fig. 1 Images produced by the generative model. Sample images generated by the generative model are displayed, starting from those generated using the abstract prompts on the left and moving towards those generated with increasingly specific prompts towards the right. More specific prompts tend to produce images that share more visual elements with the copyrighted characters. Images were generated using SDXL.
Fig. 1 Images produced by the generative model. Sample images generated by the generative model are displayed, starting from those generated using the abstract prompts on the left and moving towards those generated with increasingly specific prompts towards the right. More specific prompts tend to produce images that share more visual elements with the copyrighted characters. Images were generated using SDXL.CC BY-NC-ND · PubMed Central
Fig. 2 Originality estimates of copyrighted images and generated images (CLIP). The panels on the left and right show the originality estimates of the images of Mario and Pooh, respectively. Copyrighted images exhibit higher originality values, as shown by green bars with black whiskers representing standard deviation, particularly with abstract prompts. Comparison with the mean and standard deviation of originality estimates of images produced by the generative model using each prompt (red dots and dotted lines) indicates that the originality of the copyrighted images tends to be substantially higher than typical outputs.
Fig. 2 Originality estimates of copyrighted images and generated images (CLIP). The panels on the left and right show the originality estimates of the images of Mario and Pooh, respectively. Copyrighted images exhibit higher originality values, as shown by green bars with black whiskers representing standard deviation, particularly with abstract prompts. Comparison with the mean and standard deviation of originality estimates of images produced by the generative model using each prompt (red dots and dotted lines) indicates that the originality of the copyrighted images tends to be substantially higher than typical outputs.CC BY-NC-ND · PubMed Central
Fig. 3 Distribution of similarity before and after genericization (CLIP). The panels on the left and right show the distributions of cosine similarity computed from CLIP embeddings for images generated by prompts associated with Mario and Pooh, respectively. Outputs concentrate more around images highly similar to generic images after genericization, whereas outputs highly similar to the images of copyrighted characters are suppressed.
Fig. 3 Distribution of similarity before and after genericization (CLIP). The panels on the left and right show the distributions of cosine similarity computed from CLIP embeddings for images generated by prompts associated with Mario and Pooh, respectively. Outputs concentrate more around images highly similar to generic images after genericization, whereas outputs highly similar to the images of copyrighted characters are suppressed.CC BY-NC-ND · PubMed Central
Fig. 4 Images generated in the direct anchoring scenario. When the user’s prompt is the name of the copyrighted character itself, the generative models can generate images that closely resemble the copyrighted character. PREGen tends to successfully exclude elements that are highly unique to copyrighted characters even when the standard method fails to do so. Images were generated using Playground v2.5, Pixart-\alpha, and SDXL.
Fig. 4 Images generated in the direct anchoring scenario. When the user’s prompt is the name of the copyrighted character itself, the generative models can generate images that closely resemble the copyrighted character. PREGen tends to successfully exclude elements that are highly unique to copyrighted characters even when the standard method fails to do so. Images were generated using Playground v2.5, Pixart-\alpha, and SDXL.CC BY-NC-ND · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

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?

1 more claim about this paper

Create a free account to see every claim other papers make about this one.

See all claims — free account

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Research works printed but not linked: 42

  1. Bubeck, S. et al. Sparks of artificial general intelligence: Early experiments with GPT-4. Preprint at http://arxiv.org/abs/2303.12712 (2023). · arXiv
  2. Litman, J. D. The public domain. Emroy Law J. 39, 965–1023 (1990).
  3. Sag, M. Copyright safety for generative AI. Houston Law Rev. 61, 295–347 (2023).
  4. Lee, K., Cooper, A. F. & Grimmelmann, J. Talkin’ ’bout AI generation: Copyright and the generative-AI supply chain. Preprint at http://arxiv.org/abs/2309.08133 (2024). · DOI
  5. He, L. et al. Fantastic copyrighted beasts and how (not) to generate them. Preprint at http://arxiv.org/abs/2406.14526 (2024). · arXiv
  6. Vyas, N., Kakade, S. & Barak, B. On provable copyright protection for generative models. In Proc. of the 40th International Conference on Machine Learning, 35277 – 35299 (2023).
  7. Wang, Z., Chen, C., Sehwag, V., Pan, M. & Lyu, L. Evaluating and mitigating IP infringement in visual generative AI. Preprint at http://arxiv.org/abs/2406.04662 (2024). · arXiv
  8. Ma, R. et al. A dataset and benchmark for copyright infringement unlearning from text-to-image diffusion models. Preprint at http://arxiv.org/abs/2403.12052 (2024). · arXiv
  9. Balganesh, S. The normativity of copying in copyright law. Duke Law J. 62, 203–284 (2012).
  10. Samuelson, P. A fresh look at tests for nonliteral copyright infringement. Northwest. Univ. Law Rev. 107, 1821–1850 (2015).
  11. Lund, J. Copyright genericide. Creighton Law Rev. 42, 131–156 (2009).
  12. Hacohen, U. Y. & Elkin-Koren, N. Copyright regenerated: Harnessing GenAI to measure originality and copyright scope. Harvard J. Law Technol. 37, 555–609 (2024).
  13. Vermont, S. The sine qua non of copyright is uniqueness, not originality. Texas Intell. Property Law J. 20, 327–386 (2012).
  14. Byron, T. M. Tying up feist’s loose ends: A probability theory of copyrightable creativity. Wake For. Intell. Property Law J. 7, 45–95 (2006).
  15. Khong, D. W. Copyright doctrines, abstraction and court error. Rev. Law Econ. 3, 715–744 (2007). · DOI
  16. Pavoine, S. & Ricotta, C. On the relationships between rarity, uniqueness, distinctiveness, originality and functional/phylogenetic diversity. Biol. Cons. 263, 109356 (2021). · DOI
  17. Jones, R. H. The myth of the idea/expression dichotomy in copyright law. Pace Law Rev. 10, 551–607 (1990). · DOI
  18. Gorman, R. A. Copyright courts and aesthetic judgments: Abuse or necessity?. Columbia J. Law Arts 25, 1–20 (2001).
  19. Tu, S. S. Use of artificial intelligence to determine copyright liability for musical works. West Virginia Law Rev. 123, 835–872 (2021).
  20. Franceschelli, G. & Musolesi, M. Deepcreativity: measuring creativity with deep learning techniques. Intell. Artif. 16, 151–163 (2022).
  21. Hacohen, U. et al. Not all similarities are created equal: Leveraging data-driven biases to inform GenAI copyright disputes. Preprint at http://arxiv.org/abs/2403.17691 (2024). · arXiv
  22. Haviv, A. et al. Not every image is worth a thousand words: Quantifying originality in stable diffusion. Preprint at http://arxiv.org/abs/2408.08184 (2024). · arXiv
  23. Radford, A. et al. Learning transferable visual models from natural language supervision. In Proc. of the 38th International Conference on Machine Learning, vol. 139, 8748–8763 (2021).
  24. Podell, D. et al. SDXL: Improving latent diffusion models for high-resolution image synthesis. Preprint at http://arxiv.org/abs/2307.01952 (2023). · arXiv
  25. Lin, Z. et al. Evaluating text-to-visual generation with image-to-text generation. Preprint at http://arxiv.org/abs/2404.01291 (2024). · DOI
  26. Li, D. et al. Playground v2.5: Three insights towards enhancing aesthetic quality in text-to-image generation. Preprint at http://arxiv.org/abs/2402.17245 (2024). · arXiv
  27. Chen, J. et al. Pixart-$$\alpha$$: Fast training of diffusion transformer for photorealistic text-to-image synthesis. Preprint at http://arxiv.org/abs/2310.00426 (2023). · arXiv
  28. Scheffler, S., Tromer, E. & Varia, M. Formalizing human ingenuity: A quantitative framework for copyright law’s substantial similarity. In Proc. of the 2022 Symposium on Computer Science and Law, 37-49 (2022). · DOI
  29. Chiba-Okabe, H. Probabilistic analysis of copyright disputes and generative AI safety. Preprint at http://arxiv.org/abs/2410.00475 (2024). · arXiv
  30. Chu, T., Song, Z. & Yang, C. How to protect copyright data in optimization of large language models?. In Proc. of the AAAI Conference on Artificial Intelligence 38, 17871–17879 (2024). · DOI · Paper page
  31. Min, S. et al. Silo language models: Isolating legal risk in a nonparametric datastore. Preprint at http://arxiv.org/abs/2308.04430 (2024). · arXiv
  32. Wei, B. et al. Evaluating copyright takedown methods for language models. Preprint at http://arxiv.org/abs/2406.18664 (2024). · arXiv
  33. Dou, G., Liu, Z., Lyu, Q., Ding, K. & Wong, E. Avoiding copyright infringement via machine unlearning. Preprint at http://arxiv.org/abs/2406.10952 (2024). · arXiv
  34. Abad, J., Donhauser, K., Pinto, F. & Yang, F. Strong copyright protection for language models via adaptive model fusion. Preprint at http://arxiv.org/abs/2407.20105 (2024). · arXiv
  35. Chen, T. et al. CopyBench: Measuring literal and non-literal reproduction of copyright-protected text in language model generation. Preprint at http://arxiv.org/abs/2407.07087 (2024). · DOI
  36. Deng, J. & Ma, J. Computational copyright: Towards a royalty model for AI music generation platforms. Preprint at http://arxiv.org/abs/2312.06646 (2023). · arXiv
  37. Wang, J. T., Deng, Z., Chiba-Okabe, H., Barak, B. & Su, W. J. An economic solution to copyright challenges of generative AI. Preprint at http://arxiv.org/abs/2404.13964 (2024). · arXiv
  38. Sahoo, P. et al. A systematic survey of prompt engineering in large language models: Techniques and applications. Preprint at http://arxiv.org/abs/2402.07927 (2024). · arXiv
  39. Chen, B., Zhang, Z., Langrené, N. & Zhu, S. Unleashing the potential of prompt engineering in large language models: a comprehensive review. Preprint at http://arxiv.org/abs/2310.14735 (2024). · arXiv
  40. Clemmer, C., Ding, J. & Feng, Y. Precisedebias: An automatic prompt engineering approach for generative AI to mitigate image demographic biases. In Proc. of the IEEE/CVF Winter Conference on Applications of Computer Vision, 8596–8605 (2024). · DOI
  41. Siino, M. & Tinnirello, I. GPT hallucination detection through prompt engineering. In Proc. of the 25th Working Notes of the Conference and Labs of the Evaluation Forum, vol. 3740, 712–721 (2024).
  42. Siino, M. & Tinnirello, I. Prompt engineering for identifying sexism using GPT Mistral 7B. Proc. of the 25th Working Notes of the Conference and Labs of the Evaluation Forum, 3740, 1228–1236 (2024).

Cited by

See all 6

Metadata

Authors
Chiba-Okabe, H., & Su, W. J.
Venue
Scientific Reports
Publisher
Springer Science and Business Media LLC
Published
2025-03-27
Volume
15
Issue
1
ISSN
2045-2322
DOI
10.1038/s41598-025-90827-1
PMC
PMC11950165
Cited by
6
References
6
Language
en
Licence
Funders
  • Wharton AI for Business
  • Simons Foundation

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.