In this article, we investigate the relationship between human oversight and fairness within the evolving framework of EU AI regulatory governance. We address two core research questions: (1) How are human oversight and fairness related? and (2) To what extent does the AI Act establish a framework for human oversight that effectively supports the implementation of the various dimensions of fairness? Based on a review of interdisciplinary literature, the article identifies three normative claims linking human oversight to fairness: first, that human oversight can help mitigate bias and error in AI systems; second, that it can function as a mechanism of accountability by assigning oversight to natural persons where AI systems lack legal liability; and third, that it can introduce human empathy and contextual sensitivity into decision-making processes, enabling a substantive notion of fairness that takes into account individual circumstances. A critical analysis of the AI Act reveals that while these normative aspirations are acknowledged, the Act only partially operationalises them, leaving several aspects of fairness insufficiently supported.

Fairness in service robotics is a complex and multidimensional concept shaped by legal, social and technical considerations. As service robots increasingly operate in personal and professional domains, questions of fairness – ranging from legal certainty and anti-discrimination to user protection and algorithmic transparency – require systematic and interdisciplinary engagement. This paper develops a working definition of fairness tailored to the domain of service robotics based on a doctrinal analysis of how fairness is understood across different fields. It identifies four key dimensions essential to fair service robotics: (i) furthering legal certainty, (ii) preventing bias and discrimination, (iii) protecting users from exploitation and (iv) ensuring transparency and accountability. The paper explores how developers, policymakers and researchers can contribute to these goals. While fairness may resist universal definition, articulating its core components offers a foundation for guiding more equitable and trustworthy human–robot interactions.

An emerging field of AI, namely Fair Machine Learning (ML), aims to quantify different types of bias (also known as unfairness) exhibited in the predictions of ML algorithms, and to design new algorithms to mitigate them. Often, the definitions of bias used in the literature are observational, i.e. they use the input and output of a pre-trained algorithm to quantify a bias under concern. In reality, these definitions are often conflicting in nature and can only be deployed if either the ground truth is known or only in retrospect after deploying the algorithm. Thus, there is a gap between what we want Fair ML to achieve and what it does in a dynamic social environment. Hence, we propose an alternative dynamic mechanism, “Fair Game”, to assure fairness in the predictions of an ML algorithm and to adapt its predictions as the society interacts with the algorithm over time. “Fair Game” puts together an Auditor and a Debiasing algorithm in a loop around an ML algorithm. The “Fair Game” puts these two components in a loop by leveraging Reinforcement Learning (RL). RL algorithms interact with an environment to take decisions, which yields new observations (also known as data/feedback) from the environment and in turn, adapts future decisions. RL is already used in algorithms with pre-fixed long-term fairness goals. “Fair Game” provides a unique framework where the fairness goals can be adapted over time by only modifying the auditor and the different biases it quantifies. Thus, “Fair Game” aims to simulate the evolution of ethical and legal frameworks in the society by creating an auditor which sends feedback to a debiasing algorithm deployed around an ML system. This allows us to develop a flexible and adaptive-over-time framework to build Fair ML systems pre- and post-deployment.

As judges increasingly rely on algorithmic and AI systems in their decision-making process, concerns have been raised about their impact on the judicial duty to state reasons – a duty that is essential to the right to a fair trial and the rule of law. This paper examines whether existing legal frameworks adequately ensure this duty and its underlying normative goals in the age of automation and explores how best to safeguard this duty. Central to the analysis is the question of whether the duty to state reasons itself should be redefined and impose stricter reasoning requirements on judges using AI, requiring them to articulate both ‘pragmatic’ and technical explanations of the systems’ role in their decisions. While this approach may enhance the duty’s normative goals of legitimacy, transparency and accountability, the paper also considers the potential drawbacks of imposing such a more robust reasoning obligation in AI-assisted judicial processes.

Amid the rapidly evolving landscape of artificial intelligence (AI) regulation, a significant concern has emerged regarding the predominant focus on preemptive measures aimed at preventing or mitigating potential AI-related harms. While these preemptive measures are undeniably important, they must be complemented by effective redress mechanisms to address and remedy materialized harms. This paper highlights the crucial role of redress in empowering individuals to challenge and rectify the adverse effects of AI systems, emphasizing that access to redress is, in fact, access to justice. We critically evaluate whether current AI governance sufficiently address the need for remedies for AI-related harms, arguing that they fall short in protecting individuals’ rights. To address this gap, we outline four key steps in the redress process: (1) initiating the redress process, (2) determining appropriate avenues for redress, (3) collecting evidence to support claims and (4) receiving and responding to decisions. Each step is explored in detail, presenting distinct challenges and requirements, illustrated with real-world examples. Our findings underscore the urgent need to integrate robust redress pathways into AI governance frameworks to safeguard individual rights as AI technologies become increasingly embedded in society.

In response to the emergence of COVID in England in 2020, the government declared a suspension of face-to-face education. To deal with the cancellation of exams, then minister of education Gavin Williamson assigned the Office of Qualifications (OfQual) with devising a method for awarding grades to students based on the standards they would have met had the examinations not been cancelled. As a response, OfQual implemented a prediction matrix based on the resources available to them: this system is wildly referred to as the OfQual algorithm. However, as the points were delivered on 13 August 2020, disputes arose leading to the cancellation of the algorithm. This paper will focus on the OfQual algorithm as a particularly relevant case to highlight the tensions around the notion of fairness in the implementation of such systems. Drawing on Science and Technology Studies, I will start by opening the black box of the OfQual algorithm. This will, in turn, allow me to identify the conflicting accounts of what is considered as fair in such system and how such accounts were inscribed within this algorithm, questioning what it means for such system to be fair in contexts marked by inequalities.

This paper examines the profound challenges posed by manipulative artificial intelligence (AI) and critically evaluates the adequacy of the EU AI Act in mitigating these threats. Modern AI technologies possess the capability to influence human cognition and behaviour imperceptibly, thus endangering cognitive freedom, the fundamental right to autonomous thought. Although the EU AI Act classifies manipulative AI as an unacceptable risk and prohibits its deployment, its current framework, characterized by imprecise definitions and regulatory gaps, undermines its efficacy in holding entities accountable and safeguarding individuals. To address these deficiencies, this paper introduces an innovative analytical method that traces the origins of manipulation, enabling a systematic understanding of the harm. Central to this discussion is the expanded concept of cognitive freedom, which transcends conventional notions of thought rights to encompass protection from covert digital influence. Through illustrative case studies, such as the use of psychographic profiling in political campaigns, the paper elucidates how data-driven methodologies can be harnessed to subtly mould public perception and decision-making. The analysis further investigates digital design strategies, including targeted advertising and algorithmic curation, which constrain user autonomy and erode independent judgment. The paper advocates for a restructured EU AI Act that incorporates precise definitions, mandatory transparency and continuous oversight by independent, multidisciplinary bodies. Such enhancements would strengthen the detection and regulation of manipulative AI practices. By embedding cognitive freedom within legal protections and proposing real-time audits and comprehensive ethical assessments, this paper outlines a strategic pathway for preserving cognitive autonomy. This approach aims to mitigate the erosion of mental sovereignty and uphold the essential principles of independent thought and informed decision-making within the rapidly evolving digital landscape.

This study investigates the discriminatory impact of artificial intelligence (AI)-driven content moderation on social media platforms (SMPs), particularly in the Global South, where cultural and linguistic diversity often clash with the Western-centric AI frameworks. Platforms like Meta increasingly rely on AI algorithms to moderate vast amounts of content, but research shows that these algorithms disproportionately restrict free expression in the Global South (European Union Agency for Fundamental Rights, 2023; De Gregorio & Stremlau, 2023). This results in “over removal” – censorship of lawful content – and “slow removal,” which fails to address harmful material, both of which perpetuate inequality and hinder free speech. Through a case study on Meta, this research examines how AI-based content moderation misunderstands local contexts and systematically marginalizes users. The contributing factors include limited financial investment, inadequate language training, and political and corporate biases. The imbalance reflects power asymmetries, as governments in the Global South lack influence over platform policies. This study uses a human rights perspective to explore solutions through multistakeholder engagement, advocating for collaboration among tech companies, governments, and civil society to reform AI governance. Ultimately, it aims to inform regulatory frameworks that ensure fairer, more inclusive content moderation and protect free expression for a globally equitable digital landscape.

This article explores the human rights standards relevant to ensuring human involvement requirements in EU legislation related to automated content moderation. The opinions given by different experts and human rights bodies emphasise the human rights relevance of the way in which platforms distribute automated and human moderators in their services. EU secondary legislation establishes basic requirements for these structures that are called to be read under a human rights perspective. This article examines the justifications given for incorporating human involvement in content moderation, the different types of human involvement in content moderation, and the specific requirements for such involvement under EU secondary law. Additionally, it analyses the human rights principles concerning procedural safeguards for freedom of expression within this legal framework.

The adoption of AI is pervasive, often operating behind the scenes and influencing decisions without our explicit awareness. It impacts different aspects of our lives, from personalized recommendations to crucial determinations like hiring decisions or credit approvals. Yet, even to their developers, AI algorithms’ opacity raises concerns about fairness. The biases inherent in our data further complicate matters, as current AI systems often lack moral or logical judgment, relying solely on predictive outputs derived from learned data patterns. Efforts to address fairness in AI models face significant challenges, as different definitions of fairness can lead to conflicting outcomes. Despite attempts to mitigate biases and optimize fairness criteria, achieving a universal and satisfactory solution remains elusive. The multidimensional nature of fairness, with its roots in philosophy and evolving concepts in organizational justice, underscores the complexity of the task. Technology is inherently political, shaped by various societal factors and human biases. Recognizing this, stakeholders must engage in nuanced discussions about the types of fairness relevant in specific contexts and the potential trade-offs involved. Just as in other spheres of decision-making, navigating trade-offs is inevitable, requiring a flexible approach informed by diverse perspectives.

This study acknowledges that achieving fairness in AI is not about prescribing a singular definition or solution but adapting to evolving needs and values. Embracing ambiguity and tension in decision-making can lead to more inclusive outcomes. An interdisciplinary examination of application-specific and consensus-driven frameworks is adopted to consider fairness in AI. By evaluating factors such as application nuances, procedural frameworks, and stakeholder dynamics, this study demonstrates the framework’s expansive potential applicability in understanding and operationalizing fairness by the way of two illustrations.

Generative artificial intelligence (AI) systems, such as large language models, image synthesis tools, and audio generation engines, present remarkable possibilities for creative expression and scientific discovery but also pose pressing challenges for privacy governance. By identifying patterns in vast troves of digital data, these systems can generate hyper-realistic yet fabricated content, surface sensitive inferences about individuals and groups, and shape public discourse at an unprecedented scale. These innovations amplify privacy concerns about nonconsensual data extraction, re-identification, inferential profiling, synthetic media manipulation, algorithmic bias, and quantification. This article argues that the current U.S. legal framework, rooted in a narrowly targeted sectoral approach and overreliance on individual notice and consent, is fundamentally mismatched to address the emergent and systemic privacy harms of generative AI. It examines how the unprecedented scale, speed, and sophistication of these systems strain core assumptions of data protection law, highlighting the misalignment between AI’s societal impacts and individualistic, reactive approaches to privacy governance. The article explores distinctive privacy challenges posed by generative AI, surveys gaps in existing U.S. regulations, and outlines key elements of a new paradigm to protect individual and collective privacy rights that (1) shifts from individual to collective conceptions of privacy; (2) moves from reactive to proactive governance; and (3) reorients the goals and values of AI governance. Despite significant obstacles, it identifies potential policy levers, technical safeguards, and conceptual tools to inform a more proactive and equitable approach to governing generative AI.

This article focuses on the G7’s Hiroshima AI Process (HAIP) and its flagship document, the Hiroshima Code of Conduct, as pivotal elements in shaping global artificial intelligence (AI) governance. By conducting a comprehensive analysis of AI regulations in G7 member states, the article demonstrates a high degree of interoperability between these national frameworks and the Code of Conduct’s principles. The article proposes concrete steps to translate these principles into actionable policies at the G7 level and develops strategic adjustments to incorporate them into national standards. The article then proposes enhancements to the Code of Conduct, including the development of a common AI governance vocabulary, robust risk management frameworks, life cycle standards harmonization, effective stakeholder engagement mechanisms, specific redress mechanisms for AI harms and guidelines for government AI use to ensure democratic principles and human rights are upheld. Ultimately, this research aims to strengthen the G7’s role in leading a global AI landscape characterized by the rule of law, democracy, and human rights.

Standards complement regulation as frameworks for Artificial Intelligence governance. Within the European Union, this complementarity is laid down as the New Legislative Framework. Standards can be harmonised to provide a presumption of conformity with regulation. They draw legitimacy from the inclusion of all relevant stakeholders as well as the consensus principle although there are limitations in practice. At both European and international levels, standardisation for generative AI is still in its infancy due to standardisation following relying on a level of technical maturity. Therefore, most activity is currently seen in the policy domain. Potential directions for future generative AI standards are suggested. Generative AI drives the need for non-AI standards, too, especially in areas of digital trust and digital identity.

This paper examines the rise of monitoring schemes to coordinate supervisors and market authorities in addressing the cross-industry challenges posed by large language models’ deployment. As artificial intelligence (AI) intersects with the core mandate of market authorities dealing with financial stability, data protection, intellectual property, competition and telecommunications, effective oversight requires collaboration and information sharing. Using examples such as the Canadian Digital Regulator’s Forum, the UK’s Digital Regulation Cooperation Forum and the European Union’s AI Act implementation process, the paper illustrates how national and international institutional coordination can help operationalizing the high level principles on AI governance which are currently discussed in international fora. Ultimately, this approach aims to ensure responsible AI development while addressing risks and maximizing its societal benefits.

This paper provides an outline analysis of the evolving governance framework for artificial intelligence (AI) in Singapore. Across the Singapore government, AI solutions are being adopted in line with Singapore’s “Smart Nation Initiative” to leverage technology to make impactful changes across society and the economy. In tandem, Singaporean authorities have been assiduous to release a growing number of governance documents, which we analyse together to chart the city-state’s approach to AI governance in international comparison. The characteristics of Singapore’s AI governance approach include an emphasis on consensus-building between stakeholders (particularly government and industry but also citizens) and voluntary or “quasi” regulation, lately with an emphasis on promulgating standards (AI Standards, n.d.) and audit-like frameworks. Singaporean regulators have also been early movers (globally, and especially in the region) in the promulgation of normative instruments on AI governance including developing the world’s first AI Governance Testing Framework and Toolkit, AI Verify. The Singapore approach may be compelling for other jurisdictions in the region and around the world with an interest in a collaborative, balanced and consensual approach to governing AI outside of strict regulatory mechanisms. However, any jurisdiction adopting aspects of its evolving model would have to duly account for relevant differences in social and institutional conditions.

The Paper aims to explore whether the current ecosystem of existing and still-to-be adopted rules on artificial intelligence (AI) systems in the European Union does fully and adequately address the liability for damages caused by Generative AI system. It maps first and primarily the distinctive features and functional characteristics of Generative AI likely to impact on regulatory and legal considerations and, in particular, on determining the specific regulatory regime governing Generative AI and on tracing and allocating liability along the value chain pursuant to the AI Act. On the basis of this mapping exercise, the Paper focuses on testing the liability rules as provided for the draft Artificial Intelligence Liability Directive and the Revised Product Liability Directive and assessing their sufficiency and effectiveness in the face of Generative AI. Beyond the assessment of the rules laid down in the above-referred texts, the Paper briefly describes other liability scenarios to be explored in future works.

Large Language Models (LLMs) raises challenges that can be examined according to a normative and an epistemological approach. The normative approach, increasingly adopted by European institutions, identifies the pros and cons of technological advancement. Regarding LLMs, the main pros concern technological innovation, economic development and the achievement of social goals and values. The disadvantages mainly concern cases of risks and harms generated by means of LLMs. The epistemological approach examines how LLMs produce outputs, information, knowledge, and a representation of reality in ways that differ from those followed by human beings. To face the impact of LLMs, our paper contends that the epistemological approach should be examined as a priority: identifying risks and opportunities of LLMs also depends on considering how this form of artificial intelligence works from an epistemological point of view. To this end, our analysis compares the epistemology of LLMs with that of law, in order to highlight at least five issues in terms of: (i) qualification; (ii) reliability; (iii) pluralism and novelty; (iv) technological dependence and (v) relation to truth and accuracy. The epistemological analysis of these issues, preliminary to the normative one, lays the foundations to better frame challenges and opportunities arising from the use of LLMs.