Luciano Floridi

Applications of artificial intelligence (AI) for cybersecurity tasks are attracting greater attention from the private and the public sectors. Estimates indicate that the market for AI in cybersecurity will grow from US$1 billion in 2016 to a US$34.8 billion net worth by 2025. The latest national cybersecurity and defence strategies of several governments explicitly mention AI capabilities. At the same time, initiatives to define new standards and certification procedures to elicit users’ trust in AI are emerging on a global scale. However, trust in AI (both machine learning and neural networks) to deliver cybersecurity tasks is a doubleedged sword: it can improve substantially cybersecurity practices, but can also facilitate new forms of attacks to the AI applications themselves, which may pose severe security threats. We argue that trust in AI for cybersecurity is unwarranted and that, to reduce security risks, some form of control to ensure the deployment of ‘reliable AI’ for cybersecurity is necessary. To this end, we offer three recommendations focusing on the design, development and deployment of AI for cybersecurity.

As cyber attacks continue to escalate in terms of frequency, impact, and level of refinement so do the efforts of state actors to acquire new offensive capabilities to defend, counter or retaliate incoming attacks. Artificial Intelligence (AI) has become a key technology both for attacking and defending in cyberspace. When considered in the current regulatory vacuum this is problematic, as AI-enabled cyber conflict may escalate and threaten national security and international stability. This is why this article argues for the need to define regulation for state use of AI for defence purposes and calls on regional forums, such as NATO and the European Union, to revive efforts and prepare the ground for an initiative led by the United Nations. It concludes by offering three recommendations as key aspects to regulate. There are: legal boundaries that distinguish between legitimate and illegitimate targets and define proportionality of responses; promote testing strategies with allies to organize sparring exercises among allies to test AI-based defence tactics; monitor and enforce rules, a third-party authority with teeth, should rule on whether red lines, proportionality, responsible deployment or disclosure norms have been breached.

The automation of online social life is an urgent issue for researchers and the public alike. However, one of the most significant uses of such technologies seems to have gone largely unnoticed by the research community: religion. Focusing on Islamic Prayer Apps, which automatically post prayers from its users’ accounts, we show that even one such service is already responsible for millions of tweets daily, constituting a significant portion of Arabic-language Twitter traffic. We argue that the fact that a phenomenon of these proportions has gone unnoticed by researchers reveals an opportunity to broaden the scope of the current research agenda on online automation.

Artificial Intelligence (AI) research and regulation seek to balance the benefits of innovation against any potential harms and disruption. However, one unintended consequence of the recent surge in AI research is the potential re-orientation of AI technologies to facilitate criminal acts, term in this article AI-Crime (AIC). AIC is theoretically feasible thanks to published experiments in automating fraud targeted at social media users, as well as demonstrations of AI-driven manipulation of simulated markets. However, because AIC is still a relatively young and inherently interdisciplinary area—spanning socio-legal studies to formal science—there is little certainty of what an AIC future might look like. This article offers the first systematic, interdisciplinary literature analysis of the foreseeable threats of AIC, providing ethicists, policy-makers, and law enforcement organisations with a synthesis of the current problems, and a possible solution space.

Mobility is an essential component of life in any society, so a transformation of mobility will affect the foundations of any society, and it is hard to imagine a more profound transformation of mobility than autonomous driving. This is why understanding attitudes towards the benefits and shortcomings of autonomous vehicles means being able to address societal welfare and individual well-being more successfully. In this chapter I argue that digital technologies have made it possible to detach the journey from the trip. It seems that, in the near future, we may be increasingly able to enjoy trips rather than journeys, with more freedom to choose to travel because we want to rather than because we need to.

Research on the ethics of algorithms has grown substantially over the past decade. Alongside the exponential development and application of machine learning algorithms, new ethical problems and solutions relating to their ubiquitous use in society have been proposed. This article builds on a review of the ethics of algorithms published in 2016 (Mittelstadt et al. Big Data Soc 3(2)). The golas are to contribute to the debate on the identification and analysis of the ethical implications of algorithms, to provide an updated analysis of epistemic and normative concerns, and to offer actionable guidance for the governance of the design, development and deployment of algorithms.

An increasing number of financial services (FS) companies are adopting solutions driven by artificial intelligence (AI) to gain operational efficiencies, derive strategic insights, and improve customer engagement. However, the rate of adoption has been low, in part due to the apprehension around its complexity and self-learning capability, which makes auditability a challenge in a highly regulated industry. There is limited literature on how FS companies can implement the governance and controls specific to AI-driven solutions. AI auditing cannot be performed in a vacuum; the risks are not confined to the algorithm itself, but rather permeates the entire organization. Using the risk of unfairness as an example, this paper will introduce the overarching governance strategy and control framework to address the practical challenges in mitigating risks AI introduces. With regulatory implications and industry use cases, this framework will enable leaders to innovate with confidence.

We propose a formal framework for interpretable machine learning. Combining elements from statistical learning, causal interventionism, and decision theory, we design an idealised explanation game in which players collaborate to find the best explanation(s) for a given algorithmic prediction. Through an iterative procedure of questions and answers, the players establish a three-dimensional Pareto frontier that describes the optimal trade-offs between explanatory accuracy, simplicity, and relevance. Multiple rounds are played at different levels of abstraction, allowing the players to explore overlapping causal patterns of variable granularity and scope. We characterise the conditions under which such a game is almost surely guaranteed to converge on a (conditionally) optimal explanation surface in polynomial time, and highlight obstacles that will tend to prevent the players from advancing beyond certain explanatory thresholds. The game serves a descriptive and a normative function, establishing a conceptual space in which to analyse and compare existing proposals, as well as design new and improved solutions.

_Philosophy and Computing_ explores each of the following areas of technology: the digital revolution; the computer; the Internet and the Web; CD-ROMs and Mulitmedia; databases, textbases, and hypertexts; Artificial Intelligence; the future of computing. Luciano Floridi shows us how the relationship between philosophy and computing provokes a wide range of philosophical questions: is there a philosophy of information? What can be achieved by a classic computer? How can we define complexity? What are the limits of quantam computers? Is the Internet an intellectual space or a polluted environment? What is the paradox in the Strong Artificial Intlligence program? _Philosophy and Computing_ is essential reading for anyone wishing to fully understand both the development and history of information and communication technology as well as the philosophical issues it ultimately raises.

State legislatures in the United States have assumed a growing role in shaping artificial intelligence (AI) regulation, as federal inaction persists. This article examines the emerging landscape of state-level AI policies and the tensions that state legislators must navigate to craft effective, feasible, and forward- looking legislation. Drawing on insights from the Yale Digital Ethics Center’s 2025 Summit on State AI Legislation (SSAIL), which convened over 60 stakeholders, including lawmakers, technologists, and civil society leaders, the article addresses five key regulatory issues: open-source versus closed- source models, use-case versus model-level regulation, desired outcomes versus implementation feasibility, innovation pace versus legislative capacity, and state versus national authority. For each, it analyzes structural and political challenges and offers targeted, ranked policy recommendations for state legislators. By reframing AI governance as a potential “no-lose” scenario rather than a zero-sum game, the article argues that coordinated, context-aware state legislation can advance public interest by leveraging both innovation and regulation, helping avert the harms of unregulated AI while setting the stage for effective multi-level governance.

This article introduces the concept of “distant writing”, a novel literary practice in which authors act as designers, employing Large Language Models (LLMs) to generate narratives, while retaining creative control through precise prompting and iterative refinement. Unlike Moretti’s distant reading, which uses computational analysis to interpret large corpora of existing texts, distant writing harnesses computational tools (LLMs) to author new narratives, reshaping the literary production process. By examining theoretical frameworks and practical consequences, and running an experiment in distant writing called Encounters, this article argues that distant writing represents a significant evolution in authorship, not replacing but expanding human creativity within a design paradigm. The distinction between writing (close) and “wrAIting” (distant) reveals how LLM-assisted creativity can generate narrative possibilities previously inaccessible, transforming literature’s modal space while challenging traditional notions of authorship, creativity, and literary production. This emerging practice merits critical attention as it shapes future literary landscapes and reconfigures relationships between human creativity and artificial intelligence.

The article analyses the integration of Large Language Model (LLM)-based interfaces (editorial LLMs or eLLMs) in scholarly publishing workflows, focusing specifically on their growing role in editorial screening, manuscript preparation, and peer-review processes. It assesses the benefits eLLMs offer, including efficiency gains, improved compliance with journal guidelines, enhanced objectivity, and reduced editorial workload; and the risks, especially algorithmic biases, false positives and negatives, data privacy concerns, and potential opacity in automated decision-making. The article then offers some design recommendations for eLLMs that prioritise transparency, fairness, and user-centredness, ensuring human oversight remains integral to the editorial process. It concludes by encouraging a proactive and thoughtful engagement with these technologies to enhance scholarly publishing rather than undermine its values.

This article analyses the epistemological asymmetry between remembering and forgetting Good and Evil and their ethical implications. While philosophy has traditionally focused on the metaphysical asymmetry (the privation theory that Evil exists only as an absence of Good), this article addresses the uneven moral distribution arising from memory practices: remembering Good is morally good, whereas forgetting Good, forgetting Evil, and remembering Evil may each bear morally negative implications. The article argues that this asymmetry should be rectified through differentiated memory practices: celebration for Good and closure for Evil. Celebration involves active commemoration that perpetuates Good’s presence, while closure enables remembering Evil without causing further Evil. The article then explores practical applications of this framework to physical memory artefacts, including monuments and “cancel culture” debates, and digital memory technologies, arguing that current designs often fail to differentiate adequately between these types of “moral memory”. It concludes by suggesting some principles and requirements for designing ethical memory architectures that properly support celebration and closure, contributing to improved human flourishing through more nuanced approaches to remembering the past.

This article introduces a conjecture that formalises a fundamental trade-off between provable correctness and broad data-mapping capacity in Artificial Intelligence (AI) systems. When an AI system is engineered for deductively watertight guarantees (demonstrable certainty about the error-free nature of its outputs)—as in classical symbolic AI—its operational domain must be narrowly circumscribed and pre-structured. Conversely, a system that can input high-dimensional data to produce rich information outputs—as in contemporary generative models— necessarily relinquishes the possibility of zero-error performance, incurring an irreducible risk of errors or misclassification. By making this previously implicit trade-off explicit and open to rigorous verification, the conjecture significantly reframes both engineering ambitions and philosophical expectations for AI. After reviewing the historical motivations for this tension, the article states the conjecture in information-theoretic form and contextualises it within broader debates in epistemology, formal verification, and the philosophy of technology. It then offers an analysis of its implications and consequences, drawing on notions of underdetermination, prudent epistemic risk, and moral responsibility. The discussion clarifies how, if correct, the conjecture would help reshape evaluation standards, governance frameworks, and hybrid system design. The conclusion underscores the importance of eventually proving or refuting the inequality for the future of trustworthy AI.

Sam Altman has claimed that by the end of this year, OpenAI will be capable of "truly astonishing cognitive tasks." But what exactly does "cognition" mean in the context of artificial intelligence? As the sophistication of such technologies, our dependence on them, and the rhetoric used to sell them escalates, it's vital we grapple with this question. Earlier this year, Luciano Floridi, founding director of the Digital Ethics Center at Yale University, joined Alexandros Schismenos, post-doctoral researcher on the philosophy of artificial intelligence at the Aristotle University of Thessaloniki, and Giannis Perperidis, adjunct lecturer on the philosophy of AI at the Ionian University in Corfu, for an online conversation about the present and future of AI technology-and why solving a puzzle is still not the same as understanding it.

The digital revolution provides huge opportunities to improve private and public life, and our environments, from health care to smart cities and global warming. Unfortunately, such opportunities come with significant ethical challenges. In particular, the extensive use of increasingly more data—often personal, if not sensitive (Big Data)—the growing reliance on algorithms to analyse them in order to shape choices and to make decisions (including machine learning, AI, and robotics), and the gradual reduction of human involvement or oversight over many automatic processes, pose pressing questions about fairness, responsibility, and respect of human rights.

In this chapter, I present some ideas that I hope may help improve political thinking and practice in a mature information society. The ambition is quintessentially philosophical: trying to understand and improve the world, to the extent that each of us can contribute, in this case with some intellectual work. That is all. It is not a little, I realize, but it is not much either. It is the usual paradox: how important is a vote, or, in this case, a conceptual contribution? As much as a grain of sand on the beach: one counts for nothing, two are still nothing, but millions of grains can make a significant difference, if only because, without them, the beach would not exist. This is the relational value of aggregation. The ambition is therefore philosophical, but also aggregative, because I hope that the ideas expressed in this chapter may be useful and find some follow-up.

This chapter draws together the conclusions of the book as a whole, namely that the available typologies of group privacy, such as collective action lawsuits or the articulation of the rights of political or activist groups, are intuitively insufficient to address the landscape emerging from the new data analytic technologies. The book has demonstrated that there are multiple and often divergent perspectives on what a group is and how it should be addressed with regard to privacy, but this divergence is also an important tool for understanding which elements of the problem can be addressed using current legal and conceptual tools and which will require new approaches. The chapter outlines the authors’ contributions to the typology of group privacy concerns, then identifies the gaps and limitations that arise from a group perspective on privacy, and the conceptual and practical implications of taking the group level into account. Finally, it suggests ways forward for future research, discussion and action.

The introduction outlines the origin of the book in the concerns arising from new data analytical technologies with regard to collectives. Although their effects have so far been addressed only on the individual level, in fact profiling and machine learning technologies are directed at the group level and are used to formulate types, not tokens – they work to scale, and enable their users to target the collective as much as the individual. This means that our legal, philosophical and analytic attention to the individual may need to be adjusted, and possibly extended, in order to pay attention to the actual technological landscape unfolding before us. This book represents such an adjustment: it may be seen as an exploration of the territory that lies between ‘their privacy’ and ‘its privacy’, with regard to a given group. In this book we push the boundary towards ‘its’, and begin to think about the implications of that shift, first by identifying who must be involved in the discussion. This chapter outlines the philosophical and legal concerns that arise from considering group privacy, and sets out the challenge that this volume aims to answer.

In this chapter I identify three problems affecting the plausibility of group privacy and argue in favour of their resolution. The first problem concerns the nature of the groups in question. I shall argue that groups are neither discovered nor invented, but designed by the level of abstraction (LoA) at which a specific analysis of a social system is developed. Their design is therefore justified insofar as the purpose, guiding the choice of the LoA, is justified. This should remove the objection that groups cannot have a right to privacy because groups are mere artefacts (there are no groups, only individuals) or that, even if there are groups, it is too difficult to deal with them. The second problem concerns the possibility of attributing rights to groups. I shall argue that the same logic of attribution of a right to individuals may be used to attribute a right to a group, provided one modifies the LoA and now treats the whole group itself as an individual. This should remove the objection that, even if groups exist and are manageable, they cannot be treated as holders of rights. The third problem concerns the possibility of attributing a right to privacy to groups. I shall argue that sometimes it is the group and only the group, not its members, that is correctly identified as the correct holder of a right to privacy. This should remove the objection that privacy, as a group right, is a right held not by a group as a group but rather by the group’s members severally. The solutions of the three problems supports the thesis that an interpretation of privacy in terms of a protection of the information that constitutes an individual—both in terms of a single person and in terms of a group—is better suited than other interpretations to make sense of group privacy.

This article examines the feasibility and strategic advantages of developing open-source (OS) foundation models within the European Union (EU). Drawing upon recent developments in AI strategies globally, particularly in China, we argue that the EU’s robust regulatory framework and commitment to ethical principles uniquely position it to produce trustworthy OS foundation models. Due to stringent regulatory compliance, EU-developed OS foundation models would inherently ensure greater trustworthiness, offering a competitive and ethically aligned alternative to proprietary and foreign models. We conclude by offering concrete recommendations to realise this strategic opportunity through EU initiatives such as the AI Factories and EuroHPC infrastructure, emphasising the EU’s potential role in shaping global AI standards.

Artificial intelligence (AI) assurance is an umbrella term describing many approaches—such as impact assessment, audit, and certification procedures—used to provide evidence that an AI system is legal, ethical, and technically robust. AI assurance approaches largely focus on two overlapping categories of harms: deployment harms that emerge at, or after, the point of use, and individual harms that directly impact a person as an individual. Current approaches generally overlook upstream collective and societal harms associated with the development of systems, such as resource extraction and processing, exploitative labour practices and energy intensive model training. Thus, the scope of current AI assurance practice is insufficient for ensuring that AI is ethical in a holistic sense, i.e. in ways that are legally permissible, socially acceptable, economically viable and environmentally sustainable. This article addresses this shortcoming by arguing for a broader approach to AI assurance that is sensitive to the full scope of AI development and deployment harms. To do so, the article maps harms related to AI and highlights three examples of harmful practices that occur upstream in the AI supply chain and impact the environment, labour, and data exploitation. It then reviews assurance mechanisms used in adjacent industries to mitigate similar harms, evaluating their strengths, weaknesses, and how effectively they are being applied to AI. Finally, it provides recommendations as to how a broader approach to AI assurance can be implemented to mitigate harms more effectively across the whole AI supply chain.

This article examines potential regulatory pathways for AI in the United States following the Trump administration's 2025 revocation of the Biden-era AI Executive Order. We outline two competing governance scenarios: decentralized state-level regulation (with minimal federal oversight) and centralized federal dominance (through legislative pre-emption). We critically evaluate each model's policy implications, constitutional challenges, and practical trade-offs, particularly regarding innovation and state autonomy. We argue that AI's technological characteristics and context-dependent nature complicate achieving regulatory coherence amid competing federal and state interests. As a result, even under the Trump administration's broader deregulatory agenda, targeted federal intervention may remain necessary.