Aranovich et al. on Who Gets to See What? Information Asymmetry and the Limits of Transparency in Australia’s Medical AI Regulation

Tatiana Aranovich (Macquarie U) et al. have posted “Who Gets to See What? Information Asymmetry and the Limits of Transparency in Australia’s Medical AI Regulation” on SSRN. Here is the abstract:

Artificial intelligence (AI) is increasingly embedded in Australian healthcare, from diagnosis to patient monitoring. These technologies promise more accessible and accurate care but introduce significant safety risks. Much of this technology is also delivered through publicly funded care, which gives its transparency a public dimension alongside its clinical safety one. Transparency is widely promoted as a response to address these risks, yet little attention has been paid to whether existing medical device regulation delivers meaningful transparency (understood here as a concept distinct from explainability) for different stakeholders. This article evaluates the Australian framework through the complementary concepts of information asymmetry and meaningful transparency. Using doctrinal analysis, it maps how information flows to three principal audiences — regulators, users, and the public including independent experts — and evaluates each pathway against three criteria: accessibility, relevance and auditability. The article finds that transparency is distributed unevenly: strong to the regulator, weaker to users, and weakest to the public. It proposes a three-pillar reform agenda targeting user-facing relevance, public auditability, and the institutional translation of regulatory information into clinical practice.

Bommarito et al. on Governing AI Agents: Risk, Compliance, and Accountability in Law and Finance

Jillian Bommarito (273 Ventures) et al. have posted “Governing AI Agents: Risk, Compliance, and Accountability in Law and Finance” on SSRN. Here is the abstract:

Agentic AI systems—capable of autonomous goal pursuit, environmental perception, and iterative action—are rapidly entering legal and financial services, yet existing governance frameworks were not designed for software that acts rather than merely responds. Unlike passive AI tools, agentic systems can stray from their intended objectives and take consequential actions without human approval, creating risks that traditional compliance models often fail to address. This chapter proposes a governance framework that can be used to scale oversight requirements to match each system’s risk profile, based on factors like degree of autonomy, how long it operates, whose interests it serves, and how its objectives are set and modified. The framework situates agentic AI within a five-layer regulatory stack encompassing foundational law, professional ethics, sector-specific regulation, emerging AI-specific rules (including the EU AI Act), and voluntary assurance standards such as NIST and SOC 2.

The chapter argues that in law and finance—industries built on trust and non-delegable professional duties—human-in-the-loop and human-in-command architectures are not merely best practices but critical designs for satisfying fiduciary and regulatory obligations. To operationalize accountability without fragmenting ownership, the chapter evaluates three organizational models—centralized, federated, and embedded—and demonstrates how responsibility-assignment tools such as RACI matrices can allocate oversight duties with precision. It concludes by outlining a maturity-based adoption path that enables legal and financial institutions to adopt agentic AI while managing liability exposure and reputational risk.

Szostak et al. on Mind the Competitiveness Gap: Measuring the AI Act’s Extraterritorial Reach

Kamil Szostak (Maastricht U) et al. have posted “Mind the Competitiveness Gap: Measuring the AI Act’s Extraterritorial Reach” (Computer Law & Security Review, volume 62, 2026[10.1016/j.clsr.2026.106357]) on SSRN. Here is the abstract:

The 2024 EU Artificial Intelligence Act (AI Act) asserts broad extraterritorial scope: it applies to any AI system whose output is used in the EU, no matter where the company behind the AI is located. The AI Act reflects a wider trend of the EU extending its legislation beyond its borders. But does this mean companies actually comply with the AI Act? And do compliance levels differ between companies based inside and outside of the EU? Empirical evidence on the effectiveness of extraterritorial legislation has so far been limited, focusing mostly on privacy and consumer protection law. This paper provides such evidence by assessing demonstrated compliance with the AI Act’s prohibitions on certain AI practices under Article 5(1), for which high financial penalties started to apply in August 2025. To this end, this study analysed terms of service and acceptable use policies from 53 AI companies. Results indicate that foreign companies without EU subsidiaries rarely comply, with little improvements following the introduction of penalties. Only around 14% of such companies refer to the AI Act’s prohibitions in their documents, whereas significantly more companies with an EU establishment do so. This may result from limited (extraterritorial) enforcement and uncertainty among non-EU companies about the AI Act’s scope. The findings suggest that EU companies may face a competitive disadvantage compared with foreign firms. This concern is highly relevant to current debates on the competitiveness of EU firms and the potential need for legislative simplification.

Lee on Why AI Governance Fails Without a Legal and Scientific Definition of Consciousness

Jinho Lee (Salpida Institute Consciousness Science (SICS)) has posted “Why AI Governance Fails Without a Legal and Scientific Definition of Consciousness” on SSRN. Here is the abstract:

Artificial intelligence governance is widely recognized as one of the most urgent challenges of the twenty-first century. Governments, international organizations, and technology institutions have proposed ethical guidelines, legal frameworks, and safety principles intended to regulate increasingly autonomous AI systems. Yet despite unprecedented attention, AI governance continues to fail in practice. This paper argues that these failures are not primarily technological, political, or ethical in nature. Rather, AI governance fails because it is constructed upon an undefined and scientifically ungrounded concept of consciousness. Current legal and ethical frameworks implicitly rely on assumptions about intention, agency, responsibility, alignment, and decision-making-concepts that presuppose consciousness-while refusing to define consciousness in any measurable, operational, or legally interoperable manner. As a result, AI governance collapses into abstraction, cultural conflict, and non-enforceable norms. This paper reframes AI governance failure as a definition failure rather than a policy failure, and demonstrates why no durable AI governance system can exist without a shared legal and scientific definition of consciousness. This work functions as a diagnostic and connective document. It does not introduce new measurement models or governance mechanisms. Instead, it explains why such models are structurally required, and situates AI governance within a broader civilizational transition toward operationally defined consciousness systems already established in the Consciousness Civilization Framework (CCF) and its constitutional and operational records.

Fagan on Training Data Governance

Frank Fagan (South Texas College Law Houston) has posted “Training Data Governance” (NYU Journal of Intellectual Property & Entertainment Law, Vol. 15, pp. 227-64 (2026)) on SSRN. Here is the abstract:

As AI-generated summaries increasingly displace traditional search results, users are less likely to visit the underlying websites where content is published. This shift has sharply reduced traffic to those sites, threatening the economic viability of content creators and prompting a wave of paywalls, restrictions, and litigation. With referral-based revenue in decline, the continued supply of high-quality content faces mounting risk, precisely as generative AI has grown more dependent on such material. This tension, between innovation and sustainability, frames the central legal and policy inquiry of training data governance: how to preserve access to essential AI training inputs without undermining the incentives to produce them.

This Article examines licensing as a tool of training-data governance and focuses on the practical question of when content loss threatens model performance and reduces social welfare. The inquiry centers on whether withdrawal of high-value material is likely and whether voluntary bargaining can realistically prevent it. Where the risk of withdraw is low, additional protection is unnecessary; where the risk is substantial and bargaining fails, a narrowly tailored fallback, such as a standardized, non-exclusive license, can preserve access without disturbing fair-use doctrine or existing private arrangements. This welfare logic is implemented through a three-part test: licensing is warranted only when (1) the content has demonstrable value for AI training, (2) withdrawal is the rational market outcome absent remuneration, and (3) voluntary licensing fails due to transaction costs or bargaining frictions. Together, these conditions ensure that intervention occurs only where it improves overall welfare relative to the status quo.

Zhang et al. on The Regulatory Landscape of Generative AI in Education and the Future of Student Psychological Well-being —— A Comparative Analysis of EU and Chinese Legislation

Shengyin Zhang (Guizhou Normal U) and Jidong Wang (Northeast Normal U) have posted “The Regulatory Landscape of Generative AI in Education and the Future of Student Psychological Well-being —— A Comparative Analysis of EU and Chinese Legislation” on SSRN. Here is the abstract:

This research is intended to undertake a thorough comparative analysis of the applicability of the EU AI Act and China’s Interim Measures for the Administration of Generative AI Services within the realm of education.As generative artificial intelligence (Generative AI) becomes increasingly prevalent in educational settings, its implications for students’ learning patterns, cognitive development, and mental well-being have emerged as a central topic of global concern.By means of a meticulous examination of the texts of these two major regulations, and in conjunction with relevant expert perspectives and academic research findings, this paper systematically dissects the core disparities in their regulatory philosophies and regulatory pathways.Particular emphasis is placed on exploring potential regulatory gaps in safeguarding the mental health of students, especially minors. Additionally, this paper offers forward-looking recommendations for the refinement of future related policies.

Scattarreggia on AI-Driven Smart Contracts: Enhancing Consumer Protection or Exacerbating Consumer Protection Challenges?

Emanuele Scattarreggia (The U Sydney) has posted “AI-Driven Smart Contracts: Enhancing Consumer Protection or Exacerbating Consumer Protection Challenges?” (Journal of Law, Market & Innovation, Volume 4, No. 3, 2025. Pp. 607-636. https://doi.org/10.13135/2785-7867/12837) on SSRN. Here is the abstract:

The integration of artificial intelligence (AI) into smart contracts holds the potential to both enhance and exacerbate consumer protection challenges. Since the AI system embedded within the contract’s code enables a high degree of contractual personalisation – by tailoring the legal agreement to the unique characteristics of the targeted individual consumer, thanks to its capacity to process large amounts of personal and behavioural data in real time – it opens the door not only to scenarios of AI-powered consumer manipulation, but also to the promising opportunity of a consumer-centric AI. Such an AI would serve the consumer’s best interests by adapting the contract to their specific needs and preferences, while protecting them from – rather than exploiting – their information, cognitive, and digital vulnerabilities. This research aims to assess whether the EU legal framework – particularly the UCPD, UCTD, AI Act, GDPR, and DSA – adequately ensures that these technologies are designed and deployed with the consumer’s well-being at their core. The paper explores AI-related risks such as digital manipulation, personal data exploitation, and the black-box problem inherent in algorithmic opacity, while also addressing the liability challenge in cases of consumer harm. Ultimately, it seeks to answer whether AI-driven smart contracts can truly foster a high level of consumer protection in the AI era, by offering novel interpretations of the existing legal framework and advancing proposals for reform aligned with the fairness-by-design approach and informed by behavioural science insights.

Max on High vs. Low AGI: Ontology and Conceptual Taxonomy for Geopolitical Coherence

Antonio Max (Principal Researcher) has posted “High vs. Low AGI: Ontology and Conceptual Taxonomy for Geopolitical Coherence” on SSRN. Here is the abstract:

The rapid progression of Artificial General Intelligence (AGI) research demands conceptual tools capable of distinguishing between systems developed for open, commercial integration and those destined for sovereign, securitized deployments. Without such distinctions, risk assessments and regulatory debates collapse AGI into legacy dual-use frameworks that are ill-suited for these resources, capturing the possibility of civilian and military application but overlooking the distinct societal lineages yielded by corporate and state-grade architectures. This paper proposes a taxonomy distinguishing low-AGI and high-AGI, clarifying how commercial-economic and security-sovereign architectures can be distinguished not only by function, but by the social and political ecosystems that produce them. The taxonomy builds on international relations concepts of “high/low politics,” viewed through the lens of construal-level theory, which allows it to even capture how cooperation and conflict may coexist in the context of AGI’s emerging geopolitical stakes. By embedding AGI within power structures and securitization theory, this contribution extends dual-use discourse through an ontological taxonomy that enables more granular risk assessment and governance design-equipping policymakers and researchers to anticipate security dilemmas, institutional demands, and technical-political spillovers in the international system.

Liu et al. on China and Regulation of Artificial Intelligence

Mingwei Liu (Rutgers Management and Labor Relations New Brunswick) and Yi Sui (Renmin U China) have posted “China and Regulation of Artificial Intelligence” on SSRN. Here is the abstract:

China has emerged as a global leader in regulating artificial intelligence (AI). By analyzing major national AI-related policies and regulations, we argue that the Chinese state adopts an instrumental approach to shaping AI policies. This approach seeks to balance efficiency and security, influenced by multiple stakeholders, including tech firms, trade unions, intellectuals, journalists, workers, and the public, as well as emerging global norms of AI governance. Furthermore, China’s regulatory approach to AI is extending its global influence through AI infrastructure projects in “Belt and Road” countries, advocacy in international forums, and its participation in international technical standards bodies.

Helkala et al. on Deadbot Ethics: A Review and Analysis of Current Research

Kirsi Helkala (Norwegian Defence U College) and Carl Leichter (Norwegian Defence U College) have posted “Deadbot Ethics: A Review and Analysis of Current Research” on SSRN. Here is the abstract:

Chatbot services are a growing industry that provide digital conversation companions, where users are able to specify the appearance and behaviour of their chatbot friends. Deadbots use the same chatbot technology, but their behaviour and appearance are based on those of a real person who is now deceased.For our framework in this article, we will use the Ryan & Stahl (2020) ethical principles for AI system development and utilization. To identify the scope and extent to which current research has already studied the ethical principles related to deadbots, we start with a narrative literature review. Then we examine how the EU’s AI Act can partially mandate their ethical development and utilization. Finally, we discuss what measures could be taken to ensure that the remaining ethical principles are also mandated.