Skip to main content

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

  • Correspondence
  • Published:

Foundation models’ acceptable use policies disregard the environment and nature

Nature Machine Intelligence volume 7, pages 1771–1772 (2025)Cite this article

Subjects

Foundation models, machine learning or deep learning models that are adapted to a broad set of tasks and trained on large amounts of data1, are usually accompanied by ‘Acceptable Use Policies’ (AUPs). These specify what end uses are considered acceptable and, conversely, unacceptable. AUPs are integral to the discursive infrastructure of generative AI (GenAI)2 and contribute to establishing values and norms, with implications for how infrastructures (can) materialize. Worryingly, AUPs seldom acknowledge the harms that GenAI applications may, potentially or actually, inflict on the environment3. This is true for known direct impacts, such as greenhouse gas emissions, energy and water consumption, and material costs for model training, and even more so for system-level environmental impacts4,5.

AUPs usually include two broader categories of restrictions: content-based and use-based restrictions6. The most common content prohibitions concern: mis- and disinformation, harassment and abuse, privacy, discrimination, child harm, violence, defamation, hate, fraud, spam, and pornography6. Use restrictions target activities such as model scraping, building competing services and use in certain industries or professions6. Some prohibited uses are already regulated elsewhere or even criminalized. Others straddle moral and cultural rules, assumptions and norms. Many restrictions are vague, highly contextual and subject to change, making their enforcement difficult (for example, references to legal regulation, misleading activities or mis-/disinformation). Moreover, risk imbued in model outputs are not claimed by the developers but instead assigned to users7.

This is a preview of subscription content, access via your institution

Access options

Buy this article

  • Purchase on SpringerLink
  • Instant access to the full article PDF.

USD 39.95

Prices may be subject to local taxes which are calculated during checkout

References

  1. Bommasani, R. et al. Preprint at https://doi.org/10.48550/arXiv.2108.07258 (2022).

  2. Bennani-Taylor, S. First Monday https://doi.org/10.5210/fm.v29i2.13568 (2024).

    Article  Google Scholar 

  3. Brevini, B. Big Data Soc. https://doi.org/10.1177/2053951720935141 (2020).

    Article  Google Scholar 

  4. Galaz, V. Dark Machines: How Artificial Intelligence, Digitalization and Automation is Changing our Living Planet (Routledge, 2025).

  5. Bashir, N. et al. The climate and sustainability implications of generative AI. An MIT Exploration of Generative AI https://mit-genai.pubpub.org/pub/8ulgrckc/release/2 (2024).

  6. Klyman, K. Acceptable use policies for foundation models. In Proc. 2024 AAAI/ACM Conference on AI, Ethics, and Society (eds Das, S. et al.) 752–767 (AAAI, 2024).

  7. Edwards, L., Szpotakowski, I., Cifrodelli, G., Sangare, J. & Steward, J. Zenodo https://doi.org/10.5281/zenodo.11276104 (2024).

  8. Star, S. L. in Distributed Artificial Intelligence Vol. 2 (ed. Huhns, M. N.) 37–54 (Morgan Kaufmann, 1989).

  9. Hirvonen, N., Jylhä, V., Lao, Y. & Larsson, S. J. Assoc. Inf. Sci. Tech. 75, 1152–1165 (2024).

    Article  Google Scholar 

  10. Haider, J., Rödl, M. & White, J. Big Data Soc. https://doi.org/10.1177/20539517251365226 (2025).

    Article  Google Scholar 

Download references

Author information

Authors and Affiliations

  1. Swedish School of Library and Information Science, University of Borås, Borås, Sweden

    Björn Ekström & Jutta Haider

  2. Department of Arts and Cultural Sciences, Lund University, Lund, Sweden

    Lisa Engström

Authors
  1. Björn Ekström
    View author publications

    Search author on:PubMed Google Scholar

  2. Lisa Engström
    View author publications

    Search author on:PubMed Google Scholar

  3. Jutta Haider
    View author publications

    Search author on:PubMed Google Scholar

Corresponding author

Correspondence to Björn Ekström.

Ethics declarations

Competing interests

The authors declare no competing interests.

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Ekström, B., Engström, L. & Haider, J. Foundation models’ acceptable use policies disregard the environment and nature. Nat Mach Intell 7, 1771–1772 (2025). https://doi.org/10.1038/s42256-025-01134-3

Download citation

  • Published:

  • Version of record:

  • Issue date:

  • DOI: https://doi.org/10.1038/s42256-025-01134-3

Search

Advanced search

Quick links

Nature Briefing Anthropocene

Sign up for the Nature Briefing: Anthropocene newsletter — what matters in anthropocene research, free to your inbox weekly.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing: Anthropocene