Position: EU AI Act's Research Exemptions Can Break the Publication Norms of Major AI Conferences

Alina Wernick; Kristof Meding

REVIEW 3 major objections 4 minor 56 references

Position: EU AI Act's Research Exemptions Can Break the Publication Norms of Major AI Conferences

T0 review · 3 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read The EU AI Act's research exemptions fail to cover the standard machine-learning practice of publishing models and code, so conference releases can turn academic researchers into regulated providers.

desk verdict A clear, well-structured position paper that puts a real question on the table—whether conference artifact release can void the AI Act's research exemptions—but the key interpretive step on 'commercial activity' needs more support than the paper gives. read the letter →

arxiv 2506.03218 v2 pith:LEN67EZY submitted 2025-06-03 cs.CY cs.AIcs.LG

Alina Wernick , Kristof Meding This is my paper

classification cs.CYcs.AIcs.LG

keywords EUAIActresearchexemptionplacingonthemarketprovidergeneral-purposeacademicpublishingmachinelearningreproducibilitylegaluncertainty

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper argues that the EU AI Act's exemptions for scientific research fail to cover a core practice of machine learning: releasing trained models, code, and demos alongside conference papers. Because the Act defines 'placing on the market' broadly and includes free of charge supply, a researcher who uploads an AI system to a public repository may be classified as a 'provider' with the full documentation and risk-management duties of a commercial company. The authors walk through each relevant exception and show that each has a gap for the act of publication. The practical stakes are fines up to €35 million and a chilling effect on reproducibility and open science. The paper proposes that the EU revise the Act or issue an authoritative interpretation to restore legal certainty.

What carries the argument

The mechanism is a chain of statutory definitions: 'provider' (Art. 3(3)), 'placing on the market' (Art. 3(9)–(10)), 'putting into service' (Art. 3(11)), and the research exceptions that carve out Arts. 2(6), 2(8), and the GPAI-specific exceptions. The decisive interpretive move is reading 'in the course of a commercial activity' (Art. 3(10)) to include non-profit academic publication, relying on the Act's contrast between professional and non-professional use and on the European Commission's Blue Guide, which treats non-profit bodies as capable of commercial activity. The paper's roadmap pushes a researcher through three questions: whether the artifact is a regulated AI system or GPAI model, whether the researcher's actions trigger market placement, and whether any exception protects the specific publication act. Each exception fails at the publication step because the Act defines research narrowly as investigation, not as distribution of artifacts.

What would settle it

A binding interpretation from the EU legislator or a competent court stating that non-profit academic publication of AI models on public repositories does not constitute 'placing on the market' under Article 3(10) would refute the central claim. Conversely, any enforcement action in which an academic researcher is fined as a provider would confirm it.

Watch

Extended reading notes

Core claim

The central claim is that publishing an AI system during ordinary research—uploading code, weights, or an interactive demo as part of a paper—can legally count as placing the system on the EU market, and that this act falls outside the AI Act's research exceptions. The paper shows that the sole-purpose scientific research exception (Art. 2(6)) covers only internal deployment, the product-oriented R&D exception (Art. 2(8)) stops at market placement, the general-purpose AI exceptions are vague or appear only in non-binding recitals, and the open-source exception does not cover high-risk systems or systemic-risk GPAI models. Consequently, the same publication practice the community treats as the standard way to share knowledge can turn the author into a regulated provider. The paper does not claim this was the legislature's intention; it argues the Act's drafting misaligns with the publication norms of the machine-learning field.

Load-bearing premise

The paper assumes that a researcher uploading a model or code to a public repository is acting 'in the course of a commercial activity' under Article 3(10), making the upload a placement on the EU market; if a court instead reads 'commercial activity' to exclude non-profit academic sharing, the central claim collapses.

Editorial extensions

If this is right

Researchers who publish models at major machine learning conferences may have to implement the AI Act's risk-management, data-governance, and logging duties for those artifacts, even with no commercial intent.
Free downloads do not escape the Act: making a model available without charge still counts as placing it on the market, so open repositories are not automatically safe.
Publishing only model weights rather than a full system, with a statement of the model's intended purpose as research-only, could reduce but not eliminate the risk.
General-purpose AI models face the greatest uncertainty, since their systemic-risk research exception appears only in non-binding recitals.
Without revision or an official interpretation, the legal uncertainty could push researchers to withhold models or add restrictive licenses, undermining reproducibility.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

If courts adopt this reading, the AI Act's extraterritorial scope means researchers anywhere in the world who post a model accessible in the EU could become providers, turning every public release into a potential compliance event.
The paper's proposed disclaimer-based mitigation is untested; an empirical study of whether such intended-purpose statements appear in real model releases and how courts treat them would test its viability.
A parallel analysis could be applied to other EU product-harmonization rules that use 'placing on the market,' but the AI Act's broad definitions and extraterritorial reach make the exposure unusually wide.
A concrete test: if legal certainty is not restored, tracking model-release rates at major AI conferences over the Act's full applicability phases (2025–2027) could reveal whether publication practices actually change.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The paper is worth reading because it identifies a genuine gap: the standard practice of releasing models and code at NeurIPS/ICML/ICLR may not be covered by the AI Act's research exceptions, and the community hasn't really grappled with that. The authors are honest that this is a risk assessment, not a settled conclusion, which is the right posture for a position paper.

What's new is the focus on conference publication as the trigger. Prior work on the research exceptions mostly looks at internal R&D or open-source licenses; the specific claim that a free upload to GitHub or Hugging Face can constitute "placing on the market" and bring a researcher into provider territory is not something I've seen argued so directly. The paper also does a solid job for its intended audience: the primer in Section 2 is clear, the flowchart is useful, and the recommendations—publish models rather than systems, state intended purpose, use disclaimers—are practical even if legally uncertain.

The soft spot is exactly where the stress-test note points. Section 3.9 argues that "commercial activity" is broad enough to cover academic publication, but the support is thin: one paragraph asserting breadth, a reference to the Blue Guide, and a contrast with non-professional use. The Blue Guide's own factors—regularity, characteristics of the producer, intentions—could just as easily cut against treating a one-off academic release as commercial. The paper doesn't resolve that; it just states its reading. That's a real weakness, but it's not fatal, because the paper's contribution is flagging the uncertainty, not proving the interpretation. For a position paper, that might be enough, but a referee should push for deeper engagement with legal scholarship and maybe an analysis of how analogous questions have been treated under other EU product-harmonization rules.

I also note the paper deliberately excludes Article 50 transparency obligations, deployer duties, and military exceptions. Those are acknowledged limitations, not errors. The citation pattern looks fine; the paper cites the key studies on research exemptions and engages with the most in-depth one.

Overall: this is a serious, coherent piece that deserves a real referee. It's aimed at AI researchers who need to know about this risk, and legal scholars who should refine the "commercial activity" analysis. I'd bring it to a reading group and would cite it if I were writing about research norms or the AI Act's scope.

Recommendation: send it to peer review. It's not ready as-is—the central interpretive claim needs more support and the recommendations could be sharper—but it's a valuable contribution to a discussion that the community needs to have.

Referee Report

3 major / 4 minor

Summary. The paper is a position piece arguing that the EU AI Act can apply to academic AI researchers who publish models, code, and demos at major machine learning conferences. It provides a primer on the AI Act's categories, analyzes the scientific research, product-oriented research, and open-source exceptions, and concludes that the act of publishing research artifacts can constitute 'placing on the market' or 'making available on the market,' thereby triggering provider obligations. The paper offers a roadmap and a flowchart for researchers to assess their exposure, and it closes with legislative and practical recommendations, including publishing only models rather than systems and using disclaimers to limit intended purpose.

Significance. The paper addresses a timely and genuinely important question for the AI research community: whether ordinary publication practices such as releasing weights on Hugging Face or code on GitHub can bring researchers within the AI Act's provider obligations. Its strengths are a structured reading of the Act's text and recitals, explicit engagement with counterarguments, and concrete, actionable recommendations for researchers. The paper is also honest about the prevailing legal uncertainty. However, its significance is conditional: the central claim rests on a contested interpretation of 'commercial activity' in Article 3(10), and the paper does not yet provide sufficient legal authority to establish that academic publication falls within that notion. If that premise fails, the headline risk to standard NeurIPS/ICLR releases largely disappears. The paper is therefore a valuable starting point for debate, but its central conclusion needs firmer legal grounding.

major comments (3)

[Section 3.9] The load-bearing premise that a researcher's free upload of a model or code to GitHub or Hugging Face is 'making available on the market' under Article 3(10) is not established. The paper responds to the counterargument by citing the Blue Guide and asserting that 'commercial activity' has a broad meaning encompassing professional contexts, but it does not cite any case law or official interpretation that specifically addresses non-profit academic publication. The Blue Guide's own factors—regularity of supplies, characteristics of supplies, producer characteristics, and intentions—and its statement that occasional supplies by charities or hobbyists are not business-related plausibly point in the opposite direction for a typical academic release. The paper needs to engage with these factors in detail and explain why an academic repository release is not an occasional, non-business supply. Without this, the central claim that conference publication can trigger the AI Act's provider obligations is not sufficiently supported.
[Section 3.2] The paper uses the Oxford English Dictionary's definition of 'research' as the yardstick for deciding whether uploading research artifacts is part of scientific research, concluding that publication may not belong to the 'core research activity.' This is not a legitimate method of statutory interpretation for EU law: terms such as 'scientific research and development' in Article 2(6) and Article 2(8) must be interpreted autonomously and consistently across Union law, with regard to the provision's context and objectives, not by reference to a general dictionary. The paper needs a legal argument—for example, from CJEU case law on comparable research exemptions in other EU instruments, or from the AI Act's recitals and legislative history—to support the claim that publishing a model falls outside the research exception. As it stands, the dictionary-based argument is an ad hoc interpretive step that weakens the central claim.
[Section 4] The paper's primary practical recommendation—publish only models, not systems, and attach disclaimers limiting intended purpose—is internally tension-ridden. The paper itself acknowledges that this interpretation 'does not apply well to GPAI' and may not hold where risks to fundamental rights are foreseeable, yet the summary of recommendations presents the disclaimer strategy as a main mitigation without clearly separating the limited scenarios in which it might work from those in which it will not. Given that many high-profile research releases are GPAI models or are easily integrated into systems, the paper should either narrow the recommendation or explicitly state that for GPAI and high-risk contexts the disclaimer strategy is not a viable compliance route.

minor comments (4)

[Abstract and Section 2.3] The fine amount '35,000,00C' appears to be a typo for €35,000,000 (and 'C' is not a standard currency symbol); please correct the formatting throughout the manuscript.
[Section 2.2 and Appendix A] There are inconsistent article references, e.g., 'Art. (3)(1)' and 'Art. 2(63)'; the latter should be Article 3(63) because the GPAI model definition is in Article 3.
[Section 3.7] The sentence 'The legal consequences for prohibited uses of AI are triggered by placing the systems on the market, putting them into service, or the use of the said systems' is ambiguous about whose 'use' triggers liability; please clarify whether this refers to the provider's own use, a deployer's use, or any use.
[Figure 2] The text refers to Figure 2 as a flowchart illustrating the exceptions, but the figure is not included in the manuscript; please ensure the final version contains it and that it is legible when printed.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular dependency found; the paper's claims are legal interpretations grounded in the AI Act's text, recitals, and external sources such as the Blue Guide, and its self-citations are peripheral.

full rationale

The paper is a legal position paper, not a mathematical derivation. Its central claim—that releasing models or code at conferences may trigger provider obligations because publication can count as placing on the market and the research exceptions are narrow—is argued from the statutory text (Arts. 2(6), 2(8), 3(9), 3(10), 3(63)), recitals, and external legal materials such as the Blue Guide [39] and Colonna [16]. No equation is fitted and no prediction is generated from fitted inputs; the derivation chain is an interpretive argument, not a formal one. The paper's self-citations [5], [9], and [52] are used for background or supporting observations (e.g., that GPAI regulation is 'rather vague', deepfake categories, fundamental-rights framing) and are not the load-bearing justification for the main conclusion. Section 3.9 addresses the strongest counterargument—whether academic publication is a 'commercial activity' under Art. 3(10)—by appealing to the Blue Guide's case-by-case factors and the Act's own text, which are external sources rather than restatements of the paper's conclusion. That interpretive step is contestable, and a court could indeed read 'commercial activity' narrowly, but contestability is a correctness risk, not circularity. No specific circular step can be quoted and exhibited, so the appropriate non-circularity score is 0.

Assumptions & free parameters 0 free parameters · 3 assumptions · 0 invented entities

This is a legal position paper, so the ledger lists interpretive premises rather than data. The central claim depends on the Blue Guide reading of 'commercial activity' and on the assertion that free-of-charge artifact sharing counts as market placement.

assumptions (3)

domain assumption The quoted text and recitals of Regulation (EU) 2024/1689 accurately represent the AI Act's binding requirements.
The paper's analysis rests on the accuracy of its quotations of Articles 2, 3, 53, 99 and Recitals 25, 97, 102, 109. If these quotes are incomplete or mistranslated, the conclusions could change.
domain assumption The European Commission's 'Blue Guide' interpretation of 'commercial activity' under product legislation applies to the AI Act's Article 3(10).
The paper argues that non-profit academic publication counts as 'in the course of a commercial activity' based on the Blue Guide's broad professional-context reading. This is a key interpretive premise, but its applicability to the AI Act is not settled.
ad hoc to paper The Oxford English Dictionary's definition of 'research' is the appropriate yardstick for whether publishing artifacts is part of scientific research.
The paper uses a dictionary definition to argue that uploading artifacts might fall outside research activity because it is not 'systematic investigation'. This premise is contestable and not a legal standard.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Position: EU AI Act's Research Exemptions Can Break the Publication Norms of Major AI Conferences." pith.science (2026). https://pith.science/paper/LEN67EZY

@misc{pith2026250603218,
  author       = {Pith},
  title        = {Pith review of: Position: EU AI Act's Research Exemptions Can Break the Publication Norms of Major AI Conferences},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LEN67EZY}},
  note         = {Machine review of arXiv:2506.03218}
}

read the original abstract

The EU has become one of the vanguards in regulating the digital age. A particularly important regulation in the Artificial Intelligence (AI) domain is the 2024 enacted EU AI Act. The AI Act specifies -- due to a risk-based approach -- various obligations for providers of AI systems. These obligations, for example, include a cascade of documentation and compliance measures, which represent a potential obstacle to science. But do these obligations also apply to AI researchers? This position paper argues that, indeed, the AI Act's obligations could apply in many more cases than the AI community is aware of. Moreover, we argue that the AI Act is drafted in a manner that may unwillingly disrupt the scientific publication practices of the AI research community, with a focus on model and system release. We contribute the following: 1. We offer a high-level roadmap for AI researchers to evaluate whether they need to comply with the AI Act 2. We explain with everyday research examples why the AI Act applies to AI research. 3. We analyse the exceptions of the AI Act's applicability AI research and offer visual tool for researchers to navigate the AI Act's complex system or research exceptions 4. We establish a position the AI Act's research exceptions fail to account for current AI research conventions, as publishing AI research may void the research exceptions of the Act. 5. We propose changes to the AI Act to provide more legal certainty for AI researchers and give two recommendations for AI researchers to reduce the risk of not complying with the AI Act. We see our paper as a starting point for a discussion between policymakers, legal scholars, and AI researchers to avoid unintended side effects of the AI Act.

Figures

Figures reproduced from arXiv: 2506.03218 by the authors.

**Figure 2.** This flowchart illustrates the (simplified) system of AI Act’s research-relevant exceptions, [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

56 extracted references · 50 canonical work pages

[1]

Global ai law and policy tracker, 2024

IAPP. Global ai law and policy tracker, 2024. URL https://iapp.org/media/pdf/ resource_center/global_ai_law_policy_tracker.pdf

work page 2024
[2]

Ai watch: Global regulatory tracker, n.d

White and Case. Ai watch: Global regulatory tracker, n.d. URL https://www.whitecase. com/insight-our-thinking/ai-watch-global-regulatory-tracker

work page
[3]

URL http://data.europa

Regulation (EU) 2024/1689 Artificial Intelligence Act, 2024. URL http://data.europa. eu/eli/reg/2024/1689/oj

work page 2024
[4]

Mapping the Potential of Explainable AI for Fairness Along the AI Lifecycle

Luca Deck, Astrid Schoemäcker, Timo Speith, Jakob Schöffer, Lena Kästner, and Niklas Kühl. Mapping the potential of explainable artificial intelligence (xai) for fairness along the ai lifecycle. arXiv preprint arXiv:2404.18736, 2024

work page Pith review arXiv 2024
[5]

It’s complicated

Kristof Meding. It’s complicated. the relationship of algorithmic fairness and non-discrimination regulations in the eu ai act.arXiv preprint arXiv:2501.12962, 2025

work page arXiv 2025
[6]

Navigating the ai frontier: European parliamentary insights on bias and regulation, preceding the ai act.Internet Policy Review, 12(4), 2023

Allessia Chiappetta. Navigating the ai frontier: European parliamentary insights on bias and regulation, preceding the ai act.Internet Policy Review, 12(4), 2023

work page 2023
[7]

Non-discrimination and the ai act.Law and Digitalization

Lucía Bosoer, Marta Cantero Gamito, and Ruth Rubio-Marin. Non-discrimination and the ai act.Law and Digitalization. Arazandi, 2023

work page 2023
[8]

A Detailed Analysis of Article 50 of the EU’s Artificial Intelligence Act, 2024

Thomas Gils. A Detailed Analysis of Article 50 of the EU’s Artificial Intelligence Act, 2024

work page 2024

Show all 56 references

[9]

What constitutes a deep fake? the blurry line between legitimate processing and manipulation under the eu ai act.arXiv preprint arXiv:2412.09961, 2024

Kristof Meding and Christoph Sorge. What constitutes a deep fake? the blurry line between legitimate processing and manipulation under the eu ai act.arXiv preprint arXiv:2412.09961, 2024

2024 arXiv
[10]

Deep fakes and the artificial intelligence act-an important signal or a missed opportunity?Policy & Internet, July 2024

Mateusz Łabuz. Deep fakes and the artificial intelligence act-an important signal or a missed opportunity?Policy & Internet, July 2024. Publisher: Wiley

2024
[11]

The fundamental rights impact assessment (fria) in the ai act: Roots, legal obligations and key elements for a model template.Computer Law & Security Review, 54: 106020, 2024

Alessandro Mantelero. The fundamental rights impact assessment (fria) in the ai act: Roots, legal obligations and key elements for a model template.Computer Law & Security Review, 54: 106020, 2024

2024
[12]

Possible harms of artificial intelligence and the eu ai act: fundamental rights and risk.Journal of Risk Research, pages 1–14, 2024

Isabel Kusche. Possible harms of artificial intelligence and the eu ai act: fundamental rights and risk.Journal of Risk Research, pages 1–14, 2024

2024
[13]

Pre-print: The human-centric approach in scientific research: the ai act and the new frontiers of research ethics

Alessandro Mantelero. Pre-print: The human-centric approach in scientific research: the ai act and the new frontiers of research ethics. InData Privacy, Data Property, and Data Sharing: An Interdisciplinary Perspective for Port-pandemic Transitional Scientific Research. Boca R...

2025
[14]

Sandra Wachter. Limitations and loopholes in the eu ai act and ai liability directives: what this means for the european union, the united states, and beyond.Yale Journal of Law and Technology, 26(3), 2024

2024
[15]

How the eu can achieve legally trustworthy ai: a response to the european commission’s proposal for an artificial intelligence act.Available at SSRN 3899991, 2021

Nathalie A Smuha, Emma Ahmed-Rengers, Adam Harkens, Wenlong Li, James MacLaren, Riccardo Piselli, and Karen Yeung. How the eu can achieve legally trustworthy ai: a response to the european commission’s proposal for an artificial intelligence act.Available at SSRN 3899991, 2021

2021
[16]

Liane Colonna. The ai act’s research exemption: A mechanism for regulatory arbitrage? In YSEC Yearbook of Socio-Economic Constitutions 2023: Law and the Governance of Artificial Intelligence, pages 51–93. Springer, 2023

2023
[17]

Proposed eu ai act—presidency compromise text: select overview and comment on the changes to the proposed regulation.AI and Ethics, 3 (2):381–387, 2023

Emre Kazim, Osman Güçlütürk, Denise Almeida, Charles Kerrigan, Elizabeth Lomas, Adriano Koshiyama, Airlie Hilliard, and Markus Trengove. Proposed eu ai act—presidency compromise text: select overview and comment on the changes to the proposed regulation.AI and Ethics, 3 (2):38...

2023
[18]

The amsterdam paper: Recommen- dations for the technical finalisation of the regulation of gpai in the ai act, 2024

Natali Helberger, Laurens Naudts, and Stanislaw Piasecki. The amsterdam paper: Recommen- dations for the technical finalisation of the regulation of gpai in the ai act, 2024. 10

2024
[19]

The ai act and emerging eu digital acquis.Overlaps, gaps and, 2022

Artur Bogucki, Alex Engler, Clément Perarnaud, and Andrea Renda. The ai act and emerging eu digital acquis.Overlaps, gaps and, 2022

2022
[20]

The eu ai act-balancing human rights and innovation through regulatory sandboxes and standardization, 2022

Katerina Yordanova. The eu ai act-balancing human rights and innovation through regulatory sandboxes and standardization, 2022

2022
[21]

Generative ai, copyright and the ai act.Available at SSRN, 2024

João Pedro Quintais. Generative ai, copyright and the ai act.Available at SSRN, 2024

2024
[22]

V oicing challenges: Gdpr and ai research.Open Research Europe, 2(126):126, 2022

Katherine Quezada-Tavarez, Lidia Dutkiewicz, and Noémie Krack. V oicing challenges: Gdpr and ai research.Open Research Europe, 2(126):126, 2022

2022
[23]

A researcher’s guide for using personal data and non-personal data surrogates: Synthetic data and data of deceased people.De Lege, 2021

Katja De Vries. A researcher’s guide for using personal data and non-personal data surrogates: Synthetic data and data of deceased people.De Lege, 2021

2021
[24]

Implications of the novel eu ai act for neurotechnologies.Neuron, 112(18):3013–3016, 2024

Christoph Bublitz, Fruzsina Molnár-Gábor, and Surjo R Soekadar. Implications of the novel eu ai act for neurotechnologies.Neuron, 112(18):3013–3016, 2024

2024
[25]

The eu ai act: Implications for us health care, 2024

Sebastian Porsdam Mann, I Glenn Cohen, and Timo Minssen. The eu ai act: Implications for us health care, 2024

2024
[26]

Regulating chatgpt and other large generative ai models

Philipp Hacker, Andreas Engel, and Marco Mauer. Regulating chatgpt and other large generative ai models. InProceedings of the 2023 ACM Conference on Fairness, Accountability, and Transparency, pages 1112–1123, 2023

2023
[27]

Robustness and cybersecurity in the eu artificial intelligence act

Henrik Nolte, Miriam Rateike, and Michèle Finck. Robustness and cybersecurity in the eu artificial intelligence act. InGenerative AI and Law (GenLaw ’24) Workshop at 41 International Conference on Machine Learning, Vienna, Austria, 2024

2024
[28]

Labeled faces in the wild: A database forstudying face recognition in unconstrained environments

Gary B Huang, Marwan Mattar, Tamara Berg, and Eric Learned-Miller. Labeled faces in the wild: A database forstudying face recognition in unconstrained environments. InWorkshop on faces in’Real-Life’Images: detection, alignment, and recognition, 2008

2008
[29]

Facial recognition technology can expose political orientation from naturalistic facial images.Scientific reports, 11(1):100, 2021

Michal Kosinski. Facial recognition technology can expose political orientation from naturalistic facial images.Scientific reports, 11(1):100, 2021

2021
[30]

Practical and ethical challenges of large language models in education: A systematic scoping review.British Journal of Educational Technology, 55(1):90–112, 2024

Lixiang Yan, Lele Sha, Linxuan Zhao, Yuheng Li, Roberto Martinez-Maldonado, Guanliang Chen, Xinyu Li, Yueqiao Jin, and Dragan Gaševi´c. Practical and ethical challenges of large language models in education: A systematic scoping review.British Journal of Educational Technology...

2024
[31]

Critical infrastructure protection: Gener- ative ai, challenges, and opportunities.arXiv preprint arXiv:2405.04874, 2024

Yagmur Yigit, Mohamed Amine Ferrag, Iqbal H Sarker, Leandros A Maglaras, Christos Chrysoulas, Naghmeh Moradpoor, and Helge Janicke. Critical infrastructure protection: Gener- ative ai, challenges, and opportunities.arXiv preprint arXiv:2405.04874, 2024

2024 arXiv
[32]

Artificial intelligence in human resources management: Challenges and a path forward.California Management Review, 61(4): 15–42, 2019

Prasanna Tambe, Peter Cappelli, and Valery Yakubovich. Artificial intelligence in human resources management: Challenges and a path forward.California Management Review, 61(4): 15–42, 2019

2019
[33]

Language models are few-shot learners.Advances in neural information processing systems, 33:1877–1901, 2020

Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. Language models are few-shot learners.Advances in neural information processing systems, 33:1877–1901, 2020

1901
[34]

Gemini: a family of highly capable multimodal models.arXiv preprint arXiv:2312.11805, 2023

Gemini Team, Rohan Anil, Sebastian Borgeaud, Jean-Baptiste Alayrac, Jiahui Yu, Radu Soricut, Johan Schalkwyk, Andrew M Dai, Anja Hauth, Katie Millican, et al. Gemini: a family of highly capable multimodal models.arXiv preprint arXiv:2312.11805, 2023

2023 arXiv
[35]

The llama 3 herd of models.arXiv preprint arXiv:2407.21783, 2024

Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Amy Yang, Angela Fan, et al. The llama 3 herd of models.arXiv preprint arXiv:2407.21783, 2024

2024 arXiv
[36]

Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning.arXiv preprint arXiv:2501.12948, 2025

Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Ruoyu Zhang, Runxin Xu, Qihao Zhu, Shirong Ma, Peiyi Wang, Xiao Bi, et al. Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning.arXiv preprint arXiv:2501.12948, 2025

2025 arXiv
[37]

The falcon series of open language models.arXiv preprint arXiv:2311.16867, 2023

Ebtesam Almazrouei, Hamza Alobeidli, Abdulaziz Alshamsi, Alessandro Cappelli, Ruxan- dra Cojocaru, Mérouane Debbah, Étienne Goffinet, Daniel Hesslow, Julien Launay, Quentin Malartic, et al. The falcon series of open language models.arXiv preprint arXiv:2311.16867, 2023

2023 arXiv
[38]

Bloom: A 176b-parameter open-access multilingual language model, 2023

Teven Le Scao, Angela Fan, Christopher Akiki, Ellie Pavlick, Suzana Ili ´c, Daniel Hesslow, Roman Castagné, Alexandra Sasha Luccioni, François Yvon, Matthias Gallé, et al. Bloom: A 176b-parameter open-access multilingual language model, 2023. 11

2023
[39]

The European Union, 2022/C

European Commission.The ‘Blue Guide ´ on the implementation of EU product rules 2022, volume 65. The European Union, 2022/C

2022
[40]

Non-discrimination law in europe, a primer

Frederik Zuiderveen Borgesius, Philipp Hacker, Nina Baranowska, and Alessandro Fabris. Non-discrimination law in europe, a primer. introducing european non-discrimination law to non-lawyers.Introducing European non-discrimination law to non-lawyers (April 7, 2024), 2024

2024
[41]

European research development and deployment of ai, 2025

European Commission. European research development and deployment of ai, 2025. URL https://digital-strategy.ec.europa.eu/en/policies/european-ai-research

2025
[42]

European Commission. Commission staff working document impact assessment accompanying the proposal for a regulation of the european parliament and of the council laying down harmonised rules on artificial intelligence (artificial intelligence aact) and amending certain union l...

2021
[43]

Proposal for a regulation laying down harmonized rules on artificial intelligence (artificial intelligence act) and amending certain union legislative acts, four columns, 2024

European Parliament and Council. Proposal for a regulation laying down harmonized rules on artificial intelligence (artificial intelligence act) and amending certain union legislative acts, four columns, 2024. URL https://artificialintelligenceact.eu/wp-content/ uploads/2024/0...

2024
[44]

Improving reproducibility in machine learning research (a report from the neurips 2019 reproducibility program).Journal of machine learning research, 22(164):1–20, 2021

Joelle Pineau, Philippe Vincent-Lamarre, Koustuv Sinha, Vincent Larivière, Alina Beygelzimer, Florence d’Alché Buc, Emily Fox, and Hugo Larochelle. Improving reproducibility in machine learning research (a report from the neurips 2019 reproducibility program).Journal of machin...

2019
[45]

Reviving purpose limitation and data minimisation in data- driven systems.arXiv preprint arXiv:2101.06203, 2021

Asia J Biega and Michèle Finck. Reviving purpose limitation and data minimisation in data- driven systems.arXiv preprint arXiv:2101.06203, 2021

2021 arXiv
[46]

Regulatory sandboxes in the ai act: reconciling innovation and safety?Law, Innovation and Technology, 15(2):357–389, 2023

Thomas Buocz, Sebastian Pfotenhauer, and Iris Eisenberger. Regulatory sandboxes in the ai act: reconciling innovation and safety?Law, Innovation and Technology, 15(2):357–389, 2023

2023
[47]

The law of recitals in european community legislation

Tadas Klimas and Jurate Vaiciukaite. The law of recitals in european community legislation. ILSA J. Int’l & Comp. L., 15:61, 2008

2008
[48]

License usage and changes: a large-scale study on github

Christopher Vendome, Gabriele Bavota, Massimiliano Di Penta, Mario Linares-Vásquez, Daniel German, and Denys Poshyvanyk. License usage and changes: a large-scale study on github. Empirical Software Engineering, 22:1537–1577, 2017

2017
[49]

Open source license inconsistencies on github.ACM Transactions on Software Engineering and Methodology, 32 (5):1–23, 2023

Thomas Wolter, Ann Barcomb, Dirk Riehle, and Nikolay Harutyunyan. Open source license inconsistencies on github.ACM Transactions on Software Engineering and Methodology, 32 (5):1–23, 2023

2023
[50]

Charter of fundamental rights of european union, 2000

CFREU. Charter of fundamental rights of european union, 2000. URL https://www. europarl.europa.eu/charter/pdf/text_en.pdf

2000
[51]

The eu ai act: a medley of product safety and fundamental rights?Robert Schuman Centre for Advanced Studies Research Paper, (2023/59), 2023

Marco Almada and Nicolas Petit. The eu ai act: a medley of product safety and fundamental rights?Robert Schuman Centre for Advanced Studies Research Paper, (2023/59), 2023

2023
[52]

timeless way

Alina Wernick. Impact assessment as a legal design pattern—a “timeless way” of managing future risks?Digital Society, 3(2):29, 2024

2024
[53]

The future of european competitiveness

Mario Draghi. The future of european competitiveness. part a. Technical report, European Commission, 2024. URL https://commission.europa.eu/document/ download/97e481fd-2dc3-412d-be4c-f152a8232961_en?filename=The%20future% 20of%20European%20competitiveness%20_%20A%20competitive...

2024
[54]

Questionnaire to the commissioner-designate henna virkkunen executive vice-president for tech sovereignty, security and democracy, 2024

European Parliament. Questionnaire to the commissioner-designate henna virkkunen executive vice-president for tech sovereignty, security and democracy, 2024. URL https://hearings.elections.europa.eu/documents/virkkunen/virkkunen_ writtenquestionsandanswers_en.pdf

2024
[55]

Anatomy of a fall: On the anticipated withdrawal of the ai liability directive proposal, 2025

Cristina Frattone. Anatomy of a fall: On the anticipated withdrawal of the ai liability directive proposal, 2025. URL https://verfassungsblog.de/ anatomy-of-a-fall-aiact-aild-pld/

2025
[56]

Forbidden knowledge in machine learning reflections on the limits of research and publication.AI & Society, 36(3):767–781, 2021

Thilo Hagendorff. Forbidden knowledge in machine learning reflections on the limits of research and publication.AI & Society, 36(3):767–781, 2021. 12 A Models vs. Systems in the AI Act The AI Act differentiates between AI systems and models. The AI Act’s definition of an AIsys...

2021

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.

[1] [1]

Global ai law and policy tracker, 2024

IAPP. Global ai law and policy tracker, 2024. URL https://iapp.org/media/pdf/ resource_center/global_ai_law_policy_tracker.pdf

work page 2024

[2] [2]

Ai watch: Global regulatory tracker, n.d

White and Case. Ai watch: Global regulatory tracker, n.d. URL https://www.whitecase. com/insight-our-thinking/ai-watch-global-regulatory-tracker

work page

[3] [3]

URL http://data.europa

Regulation (EU) 2024/1689 Artificial Intelligence Act, 2024. URL http://data.europa. eu/eli/reg/2024/1689/oj

work page 2024

[4] [4]

Mapping the Potential of Explainable AI for Fairness Along the AI Lifecycle

Luca Deck, Astrid Schoemäcker, Timo Speith, Jakob Schöffer, Lena Kästner, and Niklas Kühl. Mapping the potential of explainable artificial intelligence (xai) for fairness along the ai lifecycle. arXiv preprint arXiv:2404.18736, 2024

work page Pith review arXiv 2024

[5] [5]

It’s complicated

Kristof Meding. It’s complicated. the relationship of algorithmic fairness and non-discrimination regulations in the eu ai act.arXiv preprint arXiv:2501.12962, 2025

work page arXiv 2025

[6] [6]

Navigating the ai frontier: European parliamentary insights on bias and regulation, preceding the ai act.Internet Policy Review, 12(4), 2023

Allessia Chiappetta. Navigating the ai frontier: European parliamentary insights on bias and regulation, preceding the ai act.Internet Policy Review, 12(4), 2023

work page 2023

[7] [7]

Non-discrimination and the ai act.Law and Digitalization

Lucía Bosoer, Marta Cantero Gamito, and Ruth Rubio-Marin. Non-discrimination and the ai act.Law and Digitalization. Arazandi, 2023

work page 2023

[8] [8]

A Detailed Analysis of Article 50 of the EU’s Artificial Intelligence Act, 2024

Thomas Gils. A Detailed Analysis of Article 50 of the EU’s Artificial Intelligence Act, 2024

work page 2024