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:

Why the European AI Act transparency obligation is insufficient

Nature Machine Intelligence volume 5, pages 559–560 (2023)Cite this article

Subjects

With the development of the AI Act, the EU is making the first and globally most ambitious attempt to regulate AI. However, the proposed AI Act, which uses a risk-based taxonomy for AI regulation, encounters difficulties when applied to general-purpose large language models (LLMs) and likely underestimates the risks posed by these new AI models. This correspondence warns that the AI Act must evolve further to mitigate such risks.

A main challenge is that LLMs such as ChatGPT generate unverified information and produce fictitious content with confidence. For example, ChatGPT can generate pertinent, but nonexistent, academic reading lists1. Data scientists describe such effects as being caused by “hallucination”2 and explain that they arise because LLMs function like “stochastic parrots”3. Such hallucination occurs when LLMs generate text based on their internal logic or patterns, rather than the true context, leading to confidently stated but unjustified and unverified deceptive responses. LLMs are called stochastic parrots because they repeat training data or their patterns, rather than exhibiting actual understanding or reasoning.

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

Access options

Buy this article

  • Purchase on Springer Link
  • Instant access to full article PDF
Buy now

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

References

  1. Hillier, M. TECHE https://go.nature.com/4234Yz7 (2023).

  2. Ji, Z. et al. ACM Comput. Surv. 55, 248 (2022).

    Google Scholar 

  3. Bender, E. M., Gebru, T., McMillan-Major, A. & Shmitchell, S. in Proc. 2021 ACM Conference on Fairness, Accountability, and Transparency 610–623 (ACM, 2021).

  4. Kan, M. PC Magazine https://go.nature.com/424b8iu (2023).

  5. van Bekkum, M. & Zuiderveen Borgesius, F. Comput. Law Secur. Rev. 48, 105770 (2023).

    Article  Google Scholar 

  6. Edwards, L. Expert Explainer: The EU AI Act Proposal https://go.nature.com/3IAWXdS (Ada Lovelace Institute, 2022).

  7. Hacker, P., Cordes, J. & Rochon, J. SSRN https://doi.org/10.2139/ssrn.4316944 (2022).

    Article  Google Scholar 

  8. Sun, T., Shao, Y., Qian, H., Huang, X. & Qiu, X. in Proc. 39th International Conference on Machine Learning 20841–20855 (PMLR, 2022).

  9. Hacker, P., Engel, A. & List, T. VerfBlog https://doi.org/10.17176/20230120-220055-0 (2023).

    Article  Google Scholar 

Download references

Acknowledgements

We thank M. Kretschmer for his insights, comments and support during the development of this paper. Our discussions have had a pivotal role in the formation of this paper. This work was supported by a Modern Law Review Scholarship.

Author information

Authors and Affiliations

  1. CREATe Centre, School of Law, University of Glasgow, Glasgow, UK

    Zihao Li

  2. Stanford Law School, Stanford University, Stanford, CA, USA

    Zihao Li

Authors
  1. Zihao Li
    View author publications

    You can also search for this author in PubMed Google Scholar

Corresponding author

Correspondence to Zihao Li.

Ethics declarations

Competing interest

The author declares no competing interests.

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Li, Z. Why the European AI Act transparency obligation is insufficient. Nat Mach Intell 5, 559–560 (2023). https://doi.org/10.1038/s42256-023-00672-y

Download citation

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1038/s42256-023-00672-y

Search

Advanced search

Quick links

Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

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