Hamdy & Hamdy on The Unseen Layers of AI: An Exploration of Poor Data Provenance in Model Training

Mohammad Hamdy (Almond Fintech) and Mona Hamdy (Independent) have posted “The Unseen Layers of AI: An Exploration of Poor Data Provenance in Model Training” on SSRN. Here is the abstract:

The opacity of AI model training presents a complex challenge with extensive implications. AI opacity not only threatens intellectual property rights. With limited visibility into the training process, users’ ability to assess output quality or biases is severely undermined, potentially leading to uninformed use, industry “monocultures,” and systemic risks. Economically, poor data provenance may exacerbate inequalities, privileging data providers in the Global North over their Global South counterparts, who face greater challenges in asserting their data rights. Additionally, it poses regulatory challenges for national authorities tasked with protecting citizens’ privacy, possibly triggering complex legal disputes and prompting risk-averse regulators to deny developers access to data. Consequently, these jurisdictions could be denied the opportunity to participate in AI development. Culturally, AI opacity hampers user assessment of model representativeness, which could threaten linguistic and cultural diversity and perpetuate the exclusion of certain cultures or groups.

This Policy Brief urges G20 countries to enhance data provenance in AI through regulatory and technological means. It provides an overview of various regulatory avenues for data provenance regulation and assesses their potential for success, highlighting the crucial role of the G20 in strengthening these standard-setting endeavors. The Policy Brief also explores the promise of emerging technologies in enhancing transparency within the AI sector and advocates for G20 support for these technologies as an additional means to promote transparency through market competition.

Toparlak on Between a Subject and an Object: Addressing the Social Valence of Robots

Rüya Tuna Toparlak (U Lucerne) has posted “Between a Subject and an Object: Addressing the Social Valence of Robots” on SSRN. Here is the abstract:

This article concentrates on the social valence of robots as a factor in contributing to our collaboration with robots and facilitating these connections. Chapter I starts with establishing the properties of social robots and what constitutes social valence. The paper describes the emerging association built on human-robot collaboration. The paper then moves on to describe the dangers of manipulation and how it can affect liability considerations. The social valence of robots causes them to push the boundaries of the traditional object-subject paradigm. The tensions this causes are inspected under Chapter II. Discussion surrounding the legal subjectivity of robots has so far differentiated these technologies mainly by autonomy, function, and sophistication. This paper aims to concentrate on the appearance of the robot and how it is experienced by the human. Humans do not interact in the same way with different robots. The paper holds the position that this should be an important consideration in how we approach regulation. For this purpose, the draft AI Act of the EU is inspected under section III for provisions that might be relevant to or affected by the social valence of robots.

Sulkowski on AI, ESG, and Law: Potential, Limitations, and Strategies Concerning Artificial Intelligence in Sustainability Reporting

Adam J. Sulkowski (Babson College) has posted “AI, ESG, and Law: Potential, Limitations, and Strategies Concerning Artificial Intelligence in Sustainability Reporting” on SSRN. Here is the abstract:

Sustainability reporting, also known as environmental, social, and governance (ESG) reporting, is the practice of publishing information on an organization’s non-financial performance, including its impacts and actions related to climate change, human rights, and diversity, equity, and inclusion (DEI). ESG reporting, however inconsistently executed and questioned by some critics, is used by management to affect perceptions and relationships with stakeholders, including investors, employees, customers, and regulators. It can also, some argue, result in better management. This article will consider the deployment of artificial intelligence (AI) in the context of sustainability reporting. As with other technologies, like blockchain, AI may, on its own, be overhyped as single factor that could bring about substantive, widespread change in ESG reporting practices and outcomes. This is because, as in other contexts, some degree of human involvement and the quality of data inputted into systems will remain critical. Voluntary standards and regulations with consequences for non-reporting and fraud will remain salient. This paper explores the potential and limitations of AI in the context of ESG reporting, and suggests strategies for managers, attorneys, and policy makers in addressing related legal issues.

Remolina on Mapping Generative AI Regulation in Finance and Bridging Regulatory Gaps

Nydia Remolina (Singapore Management U Yong Pung How Law) has posted “Mapping Generative AI Regulation in Finance and Bridging Regulatory Gaps” (Journal of Financial Transformation, Forthcoming) on SSRN. Here is the abstract:

Generative artificial intelligence (GenAI) is rapidly reshaping the financial services sector by introducing new avenues for innovation, efficiency, and profitability. GenAI systems, including models like “generative adversarial networks” (GANs) and “transformers”, can autonomously generate content such as synthetic data, trading strategies, and fraud detection insights, transforming traditional financial operations. However, these advancements come with new challenges, particularly in ensuring that GenAI is deployed ethically, securely, and in compliance with evolving regulatory frameworks. Current financial regulations, such as those governing anti-money laundering (AML), market integrity, financial consumer protection, among others, were originally designed for human-driven processes and do not fully address the complexities introduced by AI systems. While some jurisdictions, such as the E.U., Singapore, the U.S., and China, have launched AI regulatory initiatives, frameworks specifically tailored to the financial services industry are still a work in progress. This article seeks to provide an overview of the regulatory landscape while raising awareness of the gaps that financial institutions and regulators should address to bridge the gaps in the GenAI responsible adoption in the financial sector.

Oudin & Groza on The Governance of AI Companies: Reconciling Purpose with Profits

Paul Oudin (U Oxford) and Teodora Groza (Sciences Po Paris) have posted “The Governance of AI Companies: Reconciling Purpose with Profits” on SSRN. Here is the abstract:

Artificial intelligence (‘AI’) is both a critical driver of economic change and a source of potentially extreme negative externalities. For this reason, two leading AI companies, OpenAI and Anthropic, implemented customised governance structures with the double aim of addressing these externalities while remaining financially attractive to their investors. Other AI companies across the world adopted milder governance safeguards for that purpose. This paper studies these innovative governance frameworks by providing what is to the best of our knowledge the most comprehensive review of AI companies’ various governance structures available to date. It then shows that applicable rules are determinant in shaping companies’ ability to tailor their governance structure to their specific needs and examines the limitations of corporate laws in three European jurisdictions—France, Germany, and Italy—and, to a smaller extent, the US—more specifically, Delaware and Nevada—in enabling flexible governance structures that balance profit motives with public benefit objectives. Finally, it proposes recommendations for creating a new corporate form in the European Union to better support the peculiar needs of AI and other innovative companies, in line with the European Commission’s priorities for the next five years.

Mone et al. on Data Warfare and Creating a Global Legal and Regulatory Landscape: Challenges and Solutions

Varda Mone (Alliance U Law) et al. have posted “Data Warfare and Creating a Global Legal and Regulatory Landscape: Challenges and Solutions” (International Journal of Legal Information, 0; 2024 [10.1017/jli.2024.22]) on SSRN. Here is the abstract:

The world is witnessing an increase in cross-border data transfers and breaches orchestrated by State and non-State actors. Cross-border data transfers may lead to friction among States to localize or globalize data and to provide regulatory frameworks. “Data warfare” or information-war operations are often not covered under conventional rules; however, they are categorized as acts of espionage and subject to domestic regulations. As such, the operations are used to achieve a variety of objectives, including stealing sensitive information, spreading propaganda, and causing economic damage. Notable instances of the theft of sensitive information include the recent Bangladesh government website breach, exposing 50 million records, and the Unique Identification Authority of India (UIDAI) website hack. Regulating the “data war” under the existing principles of international law may be unsuccessful in creating robust international legal frameworks to address the associated challenges. These developments further accentuate the global divide between data-rich regions in the Global North, with strong data protection mechanisms (such as the GDPR and the California Privacy Rights Act), and regions in the Global South, where there is a lack of comprehensive data protection laws and regulatory regimes. This disparity underscores the urgent need for global cooperation for substantial international regulatory mechanisms. This article examines the complexities surrounding data warfare; it highlights the imperative need for establishing a robust global legal framework for data protection, delving into the concept of data war. It also acknowledges the growing influence of advanced technologies like data computing and mining and their ongoing threats to the fundamental rights of individuals associated with exposed personal data. The authors address the deficiencies in international legal provisions and advocate for a global regulatory approach to data protection as a critical means of safeguarding personal freedoms and countering the escalating threats in the digital age.

Greenleaf on EU AI Act: Brussels Effect(s) or a Race to the Bottom?

Graham Greenleaf (Macquarie U Macquarie Law (Sydney) has posted “EU AI Act: Brussels Effect(s) or a Race to the Bottom?” ((2024) 190 Privacy Laws & Business International Report 1, 3-6) on SSRN. Here is the abstract:

The expression ‘the Brussels effect’ is often used rather loosely to refer to any or all of the ways by which EU legislative standards come to be adopted in the practices of companies (or governments) in countries outside the EU (‘third party countries’).

This article considers the EU’s Artificial Intelligence Act (AI Act), and the various ways that it could have four types of ‘Brussels effects’. We need to distinguish: Extra-territorial application; De facto corporate adoption; Legislative emulation by 3rd countries; and Adoption in international agreements and standards.

The article argues that evaluation of the EU’s influence on the regulation of AI outside the EU requires all four types of ‘Brussels effect’ to be taken into account, because EU influence can take many forms. The combination of all four versions, if it is effective, is an example of the ‘race to the top’ in multi-jurisdictional regulatory standards. The article concludes that, while it is too early to assess the extent to which the EU AI Act will be another successful example of the Brussels effects, so far, the signs are promising.

Woemmel et al. on Public Attitudes Toward Algorithmic Risk Assessments In Courts: A Deliberation Experiment

Arna Woemmel (U Hamburg Business) et al. have posted “Public Attitudes Toward Algorithmic Risk Assessments In Courts: A Deliberation Experiment” on SSRN. Here is the abstract:

We study public attitudes toward algorithmic risk assessment tools in the criminal justice system using an online deliberation study with 2,358 UK participants and apply quantitative text analysis to identify key topics, biases, and sentiments underlying these attitudes. Participants were presented with a scenario about algorithmic tools used for early release decisions and then randomly assigned in groups of three to deliberate on the scenario via free-form messenger chats. The scenarios varied between subjects, but not within groups, in three algorithmic features: (i) inclusion vs. exclusion of discriminatory variables in the tool’s input data, (ii) development by private vs. public institutions, and (iii) full vs. limited judicial discretion over the tool. Prior to group deliberation, the majority approved of these tools, with particularly high approval for tools developed by public institutions or allowing full judicial discretion. However, deliberation significantly reduced approval in all treatment groups and diminished the effects of information treatments, leading to a convergence of attitudes across groups. Text analysis suggests a negativity bias in the deliberation process, with arguments against the tools (e.g., algorithmic bias) showing stronger associations with attitude changes than supportive arguments (e.g., cost savings), even though both types of arguments were equally present in the discussions. These findings highlight the malleability of stated public approval for these tools, especially when they are deliberated in greater depth.

Rizzo & Hassan on AI Risk Management in Tax Audits: A Comparative Review of the EU and US Regulatory Approaches

Amedeo Rizzo (U Oxford Law) and Giorgio Hassan (SDA Bocconi) have posted “AI Risk Management in Tax Audits: A Comparative Review of the EU and US Regulatory Approaches” on SSRN. Here is the abstract:

This paper focuses on the AI risk management framework that applies to tax authorities under the EU and US legal systems. In recent years, the development of AI has entered the field of tax administration, revolutionizing the planning and operational tasks of tax authorities. In this scenario, it is crucial that taxpayers are not unduly exposed to any risk of harm arising from the unsafe implementation of AI by tax authorities. In this regard, the EU legal framework – with the GDPR and the recent AI Act – and the US legal framework – with the recent Executive Order on the development of Safe, Secure, and Trustworthy AI – provide valuable sources of risk-based obligations that could adequately address the risks of AI in the tax domain.

On the EU side, the GDPR and AI Act have a complementary approach – a “rights-based approach” in the case of the GDPR, and a “risk-based approach” in the case of the AI Act – and an overlapping scope of application. In the field of AI risk management, the potential overlap between the GDPR and the AI Act may provide valuable indications for adapting the GDPR-based risk management framework to the realm of AI, and, at the same time, for interpreting the scope of the AI Act in light of the rights provided under the GDPR. On the US side, the risk management obligations stemming from the Executive Order on Safe AI draw from the recent developments in AI regulation in the EU, providing measures that have a similar scope to the requirements of the AI Act. From this perspective, we discuss that the EU and US approaches to AI regulation are slowly aligning and are similarly able to address the risks arising from the use of AI in the tax domain – such as, particularly, the risks concerning AI-enabled discrimination and human-AI interaction. However, both in the EU and the US, it is unclear whether the risk management framework provided by these regulations can effectively extend to tax authorities. Except for the GDPR, the AI Act and the Executive Order seem to consider tax-related AI systems at a lower risk class compared to other categories of “high-risk” or “risk-impacting” AI systems. The misalignment in the classification of tax-related AI systems could jeopardize the application of the AI risk management framework provided in these regulations, and consequently, expose taxpayers to significant risks of harm.

For this reason, we argue that the risks concerning the use of AI in tax administration, and the benefits that could derive from the adoption of a risk management framework inspired by these three regulations, should convince EU and US lawmakers to adopt a precautionary and uniform approach to the risk categorization of tax-related AI systems. Particularly, lawmakers should locate tax-related AI systems among the pool of high-risk and rights-impacting systems for the purposes of the AI Act and the Executive Order, for the better interest of taxpayers in the EU and the US.

Davies on Artificial Intelligence & FINRA Arbitration Awards: Utilizing AI and Arbitral Analytics to Uncover FINRA Arbitration Award Patterns

Ben Davies (U Calgary Law) has posted “Artificial Intelligence & FINRA Arbitration Awards: Utilizing AI and Arbitral Analytics to Uncover FINRA Arbitration Award Patterns” on SSRN. Here is the abstract:

This paper creates novel analytics and data on FINRA arbitration awards with novel artificial intelligence models fine-tuned and fed unsupervised FINRA awards to provide sentiment analysis on individual sentences and entire awards. This AI generated data, combined with prior FINRA arbitral analytics AI research, results in interesting analytics on which words, sentence structures, and awards (from 2008 to 2020) could increase the chances of a party winning a FINRA claim even if the claim is weak. To better understand AI and arbitral analytics, this paper will provide a brief history of these respective fields, ethical issues surrounding AI usage in the legal field, prior FINRA research, and short conclusion.