Asay on Artificial Code

Clark D. Asay (Brigham Young U J. Reuben Clark Law) has posted “Artificial Code” on SSRN. Here is the abstract:

Generative artificial intelligence (AI) has rapidly transformed software development, enabling the production of high-quality code at unprecedented speed and scale. Yet despite widespread litigation over the use of copyrighted works to train AI systems in fields such as music, journalism, and visual art, the software industry has thus far remained largely absent from the AI copyright wars. The explanation lies in the pervasiveness of free and open source software (FOSS), which has provided AI developers with a vast, low-risk corpus of training material while simultaneously insulating them from copyright liability. Ironically, however, the same AI technologies that have benefited from the FOSS ecosystem now threaten to undermine its legal and normative foundations.

The core issue, this Article contends, is that much AI-generated software-what this Article terms “artificial code”-is not eligible for copyright protection under prevailing U.S. law. Because copyright requires human authorship, artificial code often falls into the public domain, rendering traditional copyright-based FOSS licensing frameworks increasingly ineffective. As copyright recedes as a viable governance mechanism, software innovators are likely to turn instead to trade secrecy and patents to protect their software innovations. These alternative regimes, while legally robust, are often at odds with the openness, reciprocity, and collaborative norms that have long defined the modern software economy.

This shift has profound implications. As more software is withheld from public access or encumbered by exclusionary patent rights, the software commons risks contraction. In turn, the contraction of that commons threatens the future development of AI systems themselves. Generative models depend on continued access to large volumes of humancurated software to avoid degradation and “model collapse,” as well as inventive AI innovations. If the FOSS ecosystem erodes, AI systems may find themselves deprived of the very inputs necessary to sustain their progress.

By tracing the historical rise of FOSS, diagnosing the legal vulnerabilities introduced by artificial code, and examining the growing pull of alternative intellectual property regimes, this Article highlights a paradox at the heart of modern AI development: generative AI may be sowing the seeds of its own stagnation. The Article concludes by exploring how these concerns may arise in other creative contexts, too, as well as assessing possible responses aimed at preserving openness in software and AI innovation while accommodating the realities of AI-generated code.

Richards et al. on Why AI Preemption Would Be a Mistake: Testimony of Professor Richards before the US House Judiciary Committee

Neil M. Richards (Washington U Law) et al. have posted “Why AI Preemption Would Be a Mistake: Testimony of Professor Richards before the US House Judiciary Committee” on SSRN. Here is the abstract:

Congress has been considering pre-empting state laws that regulate AI for a variety of reasons. This comment argues that federal preemption of State AI laws would be a mistake for three reasons. First, AI is still in its early stages, and its potential harms and benefits are still uncertain. Preemption now of state laws touching AI would be reckless and leave consumers exposed to great harm. Second, state regulation, as opposed to federal regulation, has been the primary driver of sensible tech regulation over the past three decades of our Internet age. States have proven capable of adapting and reacting to these novel issues, and federal preemption now would prevent the states from being to respond and experiment to solve both current problem we know about, and future problems that have not yet become understood. Third, industry claims that state regulation stifles innovation are historically and empirically unsound. The states have proven effective at navigating technology regulation. They have done so in a manner that has helped foster the explosive growth of the American technology industry while also building out protections for, and thereby driving the confidence of, American consumers in these new technologies. Additionally, caution and measured action now does not preclude Congress’s taking up broad AI or technology bills in the future, when the dangers around these new technologies and the merits and costs of the various state approaches have become clearer. Broad preemption today would deprive our country of the kinds of state laws that will be essential to build the necessary trust that AI systems— like the internet before them—need to become widely adopted and fulfill their potential. Thus, preemption would be a danger to the kinds of broad adoption of useful AI technologies.

Velazquez et al. on The Divided States of Work Law: Regulating US Workplaces in the Age of AI

Alvin Velazquez (Indiana U Maurer Law) and Charlotte Garden (U Minnesota Law) have posted “The Divided States of Work Law: Regulating US Workplaces in the Age of AI” (in Artificial Intelligence and Labour Law: A Global Overview, M. Biasi, ed., Giappichelli/Routledge (forthcoming, 2026)) on SSRN. Here is the abstract:

United States labor policy is notable for how little it constrains most employers: unlike many jurisdictions represented in this volume, “atwill” employment is standard, and union density is low. In line with this general approach, federal lawmakers have not regulated AI or related technologies, either to plan for feared job displacements, or to limit the use of this technology to depress worker power through surveillance, discrimination, or other strategies. And while federal regulatory agencies previously took steps to bring existing law to bear in these new contexts, this approach is far from comprehensive. However, a handful of US state governments have shown greater willingness to regulate employers’ use of these emerging technologies – although their regulation occurs in the shadow of federal “preemption doctrine,” under which states are precluded from regulating some aspects of work law. This chapter explains why federal regulation of employers’ use of emerging technologies is unlikely and analyzes prospects for states to step into the breach.

Madsen et al. on The 7Vs of AI Slop: A Typology of Generative Waste

Dag Øivind Madsen (U South-Eastern Norway USN Business) and Richard W. Puyt (U Twente) have posted “The 7Vs of AI Slop: A Typology of Generative Waste” on SSRN. Here is the abstract:

The rise of generative artificial intelligence has introduced a new shorthand in digital culture: AI slop. The term refers to the flood of low-quality, machine-generated content that now saturates platforms, from generic essays and clickbait blogs to uncanny images, deepfake pornography, and synthetic video. This paper develops a typology of AI slop structured around seven dimensions-the 7Vs: Volume (scale of production), Velocity (speed of generation and circulation), Variety (range of forms and genres), Value (erosion of cultural and epistemic worth), Verification (problems of truth and trust), Visibility (algorithmic amplification), and Virality (meme-like diffusion). Together, these dimensions explain how generative waste proliferates, why it dominates attention economies, and what risks it poses to knowledge infrastructures, creative industries, and democratic publics. Situating AI slop within theories of platform capitalism, information overload, and epistemic pollution, the paper argues that slop is not a passing nuisance but a structural feature of contemporary media ecologies. The 7Vs provide a framework for analyzing generative waste across cultural, academic, commercial, political, and organizational domains, offering a foundation for future empirical and conceptual research into how societies can navigate and mitigate slop-saturated environments.

Fang on The Tragedy of the AI Data Commons: Legal Constraints and the Economics of AI Innovation

Max Fang (Indiana U Bloomington The Vincent and Elinor Ostrom Workshop Political Theory & Policy Analysis) has posted “The Tragedy of the AI Data Commons: Legal Constraints and the Economics of AI Innovation” (Texas Intellectual Property Law Journal, Vol. 34 (forthcoming, 2026).) on SSRN. Here is the abstract:

The exponential growth of artificial intelligence (AI) has brought data governance to the forefront of legal and economic debate. As global AI models hunger for massive datasets, the legal restrictions around data ownership, privacy, and intellectual property are increasingly shaping and fragmenting what kinds of data are available, how they are accessed, and who controls them. This article investigates the institutional structures governing AI data and poses the question of how privacy and intellectual property laws influence the availability and sustainability of AI training datasets. Through a comparative institutional analysis of U.S. and EU legal regimes, it employs law-and-economics methodologies and Ostrom’s theory to frame AI training data as a commons. The paper integrates doctrinal legal review with economic reasoning to evaluate how data governance institutions simultaneously protect and constrain innovation. The article introduces the concept of a “tragedy of the AI data commons,” and argues that current incentive misalignments between data protection and AI development may lead to under-contribution and eventual data scarcity. It contributes to institutional scholarship by reconceptualizing data governance as a resource allocation problem and calls for new self-governance frameworks that promote sustainable and equitable access to AI data while preserving fundamental rights.

Afek on Accuracy and Fluency: Understanding Hallucinations in Large Language Models

Ran Afek (Tel Aviv U) has posted “Accuracy and Fluency: Understanding Hallucinations in Large Language Models” on SSRN. Here is the abstract:

Large Language Models (LLMs) have demonstrated remarkable capabilities in generating fluent, contextually appropriate text, yet they are inherently prone to hallucinations, producing outputs that appear plausible but are factually inaccurate. Drawing on the findings of Kalai et al. (2025), this paper examines how hallucinations arise, not merely from corrupted training data but from the intrinsic architecture and probabilistic nature of LLMs. This study highlights how statistical patterns learned from massive datasets underpin both the strengths and limitations of language generation models. Ultimately, LLMs favor conversational fluency over strict factual precision, reflecting the probabilistic and context-driven nature of human language itself.

Yoo et al. on U.S. Artificial Intelligence Regulation During the Biden Administration

Christopher S. Yoo (U Pennsylvania Carey Law) and Alex Mueller (U Pennsylvania Carey Law) have posted “U.S. Artificial Intelligence Regulation During the Biden Administration” (Journal of Law & Economic Regulation, Vol. 17, p.7, 2024) on SSRN. Here is the abstract:

The past four years have seen a notable evolution in the United States’ approach to regulating artificial intelligence (AI). Under the leadership of President Joseph Biden, the U.S. federal government has demonstrated a much greater willingness to address risks of harm posed by AI systems proactively. Yet in so doing, it has remained committed to fostering an environment conducive to innovation, continuing to show a preference for voluntary self-regulatory measures and sector-specific oversight rather than heavy handed interventions.

This article examines the major regulatory developments that have taken place during the Biden administration, including those at the federal level – such as the recent landmark Executive Order on AI, the National Institute of Standards and Technology (NIST) AI Risk Management Framework, and actions by specific agencies – as well as the flurry of legislative activity occurring at the state level. While questions remain about the long-term trajectory of AI regulation in the U.S., a clear governance framework has taken shape over the past four years that reflects a deliberate effort to balance safeguarding of public and national interests with the goals of promoting technological advancement and economic competitiveness.

Kildeev on Artificial Intelligence is Slave of the Lamp

Adel Kildeev (Zhdanov A. Leningrad State U) has posted “Artificial Intelligence is Slave of the Lamp” (Kildeev, Adel. Artificial Intelligence is Slave of the Lamp. Scientific Platform XXI Century, Issue 6 (September 2025), pp. 3-31 Published Version. http://www.episs.ru) on SSRN. Here is the abstract:

This paper is the published version of the article “Artificial Intelligence is Slave of the Lamp” appearing in Scientific Platform XXI Century. The article offers a legal-philosophical analysis of artificial intelligence and demonstrates that AI systems cannot be considered legal subjects under any classical or modern doctrine. Building on Roman law and contemporary theories of legal personality, the paper argues that AI constitutes a special object of law-a digital analogue of the instrumentum vocale, the “speaking tool.” The article critiques proposals for electronic personhood, evaluates major regulatory approaches (EU AI Act, UN initiatives, US/China frameworks), and concludes by proposing a UN Convention on Artificial Intelligence. AI possesses no volition, intention, or legal agency: it remains a tool-a “slave of the lamp.

Halil et al. on Big is Not Bad, but Big AI Might Be: EU Merger Law and the Future of AI Competition

Defne Halil (Maastricht U Law) and Konrad Kollnig (Maastricht U) have posted “Big is Not Bad, but Big AI Might Be: EU Merger Law and the Future of AI Competition” on SSRN. Here is the abstract:

The growing dominance of a few firms in AI increases competition concerns in the EU, as US-based tech firms, specifically Microsoft, Google, Amazon, Nvidia, Meta and Apple (“Big AI”) control critical layers of the AI stack ranging from hardware to applications. Looking at 71 notable AI startup acquisitions by Big AI from 1989 to 2024, we find that, in the past, they largely escaped the EU Merger Regulation scrutiny; most transactions fell below turnover thresholds leaving potential anti-competitive effects unexamined. The CJEU’s 2024 Illumina/Grail judgment further restricts the European Commission’s ability to review non-notifiable mergers, reinforcing thestatus quo of consolidation by Big AI, and relying instead on member state law—which we analyse across all 27 EU member states. A Targeted Approach under Article 14 of the EU’s Digital Markets Act is proposed to create a special regime for Big AI, ensuring systematic review of their acquisitions to safeguard competition in Europe and expanding beyond the current proposals in literature for value-based thresholds or Article 22 EUMR call-in referrals by member states.

Shima on Ethical AI in Multilingual Institutions: Semantic Drift in Fairness Principles and the Hybrid Justice Model

Hiroki Shima (Columbia U) has posted “Ethical AI in Multilingual Institutions: Semantic Drift in Fairness Principles and the Hybrid Justice Model” on SSRN. Here is the abstract:

Multilingual public institutions such as the United Nations, the European Union, and ASEAN increasingly deploy AI systems for recruitment, service delivery, and regulatory tasks under ethics frameworks that emphasize fairness, transparency, and non-discrimination. These principles, however, are usually drafted in English and then translated into multiple legal and working languages. Research in legal and institutional translation shows that normative terms such as “fair,” “equity,” and “accountability” rarely map cleanly across languages and legal cultures, creating systematic “semantic drift” between source texts and target versions. At the same time, global AI governance instruments—from UN-system resolutions to regional guides such as the ASEAN Guide on AI Governance and Ethics and the OECD AI Principles—are increasingly used as reference points for national and organizational policy. This paper conceptualizes “semantic drift” in translated AI ethics and fairness policies as a distinct layer of risk for organizational justice and trust in AI-supported processes. Building on the Hybrid Justice Model (HJM), it treats semantic drift as a mechanism that can widen the Justice Gap Index (JGI) and generate “semantic trust drift” in the Perception Shift Index (PSI) across language communities, even when underlying models satisfy standard statistical fairness constraints. Drawing on insights from legal translation, institutional translation, and cross-cultural value research (e.g., GLOBE, Schwartz), the paper proposes a Multilingual Fairness Drift Framework that links three elements: (1) textual drift in fairness‐related concepts across languages, (2) implementation drift as policies are operationalized in local procedures and explanations, and (3) perception drift as candidates and citizens experience AI-mediated decisions. Using UN, EU, and ASEAN examples as comparative reference points, the paper illustrates how misaligned translations of fairness principles can produce divergent expectations and justice perceptions across language groups, even within a single institution. It then outlines a research agenda that uses HJM indicators—particularly μ (cultural and linguistic alignment) and PSI—to measure and govern semantic drift over time. The contribution is twofold: it reframes multilingual translation not as a technical afterthought but as a core dimension of ethical AI governance, and it extends organizational justice theory to cover the textual and semantic infrastructures through which AI ethics principles are communicated and implemented in multilingual institutions.