Connecting the dots in trustworthy Artificial Intelligence: From AI principles, ethics, and key requirements to responsible AI systems and regulation

Information Fusion, Journal Year: 2023, Volume and Issue: 99, P. 101896 - 101896

Published: June 24, 2023

Trustworthy Artificial Intelligence (AI) is based on seven technical requirements sustained over three main pillars that should be met throughout the system's entire life cycle: it (1) lawful, (2) ethical, and (3) robust, both from a social perspective. However, attaining truly trustworthy AI concerns wider vision comprises trustworthiness of all processes actors are part cycle, considers previous aspects different lenses. A more holistic contemplates four essential axes: global principles for ethical use development AI-based systems, philosophical take ethics, risk-based approach to regulation, mentioned requirements. The (human agency oversight; robustness safety; privacy data governance; transparency; diversity, non-discrimination fairness; societal environmental wellbeing; accountability) analyzed triple perspective: What each requirement is, Why needed, How can implemented in practice. On other hand, practical implement systems allows defining concept responsibility facing law, through given auditing process. Therefore, responsible system resulting notion we introduce this work, utmost necessity realized processes, subject challenges posed by regulatory sandboxes. Our multidisciplinary culminates debate diverging views published lately about future AI. reflections matter conclude regulation key reaching consensus among these views, will crucial present our society.

Language: Английский

---

Managing the race to the moon: Global policy and governance in Artificial Intelligence regulation—A contemporary overview and an analysis of socioeconomic consequences

Discover Artificial Intelligence, Journal Year: 2024, Volume and Issue: 4(1)

Published: Feb. 26, 2024

Abstract This paper delves into the complexities of global AI regulation and governance, emphasizing socio-economic repercussions rapid development. It scrutinizes challenges in creating effective governance structures amidst race, considering diverse perspectives policies. The discourse moves beyond specific corporate examples, addressing broader implications sector-wide impacts on employment, truth discernment, democratic stability. analysis focuses contrasting regulatory approaches across key regions—the United States, European Union, Asia, Africa, Americas thus highlighting variations commonalities strategies implementations. comparative study reveals intricacies hurdles formulating a cohesive policy for regulation. Central to is examination dynamic between innovation slower pace ethical standard-setting. critically evaluates advantages drawbacks shifting responsibilities government bodies private sector. In response these challenges, discussion proposes an innovative integrated model. model advocates collaborative network that blends governmental authority with industry expertise, aiming establish adaptive, responsive regulations (called “dynamic laws”) can evolve technological advancements. novel approach aims bridge gap advancements essential processes law-making.

Language: Английский

---

A Teleological Interpretation of the Definition of DeepFakes in the EU Artificial Intelligence Act—A Purpose‐Based Approach to Potential Problems With the Word “Existing”

Policy & Internet, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 15, 2025

ABSTRACT The EU Artificial Intelligence Act is the world's first attempt to holistically regulate artificial intelligence. It presents an extensive, multi‐faceted definition of deepfakes, and introduces specific safeguards against their misuses. These guardrails have potential become a global role model. AI uses concepts that leave room for interpretation, which important given constant development technology need adjustments. However, some solutions raise problem vagueness, in turn may result narrower interpretation linguistic nature, reduce scope legally permissible countermeasures. aim this study critically evaluate deepfakes contained relation word “existing” used. A narrow could potentially exclude synthetic media from transparency obligations due non‐classification these as deepfakes. Therefore, teleological provisions proposed, reinforced with elements systemicity, so built by also include do not depict any identifiable pre‐existing persons, objects, places, entities or events better reflect goals regulation, complement value‐based system Act.

Language: Английский

---

ChatGPT- versus human-generated answers to frequently asked questions about diabetes: A Turing test-inspired survey among employees of a Danish diabetes center

PLoS ONE, Journal Year: 2023, Volume and Issue: 18(8), P. e0290773 - e0290773

Published: Aug. 31, 2023

Large language models have received enormous attention recently with some studies demonstrating their potential clinical value, despite not being trained specifically for this domain. We aimed to investigate whether ChatGPT, a model optimized dialogue, can answer frequently asked questions about diabetes. conducted closed e-survey among employees of large Danish diabetes center. The study design was inspired by the Turing test and non-inferiority trials. Our survey included ten two answers each. One these written human expert, while other generated ChatGPT. Participants had task identify ChatGPT-generated answer. Data analyzed at question-level using logistic regression robust variance estimation clustering participant level. In secondary analyses, we investigated effect characteristics on outcome. A 55% margin pre-defined based precision simulations been published as part protocol before data collection began. Among 311 invited individuals, 183 participated in (59% response rate). 64% heard ChatGPT before, 19% tried it. Overall, participants could 59.5% (95% CI: 57.0, 62.0) time, which outside zone. characteristics, previous use strongest association outcome (odds ratio: 1.52 (1.16, 2.00), p = 0.003). Previous users answered 67.4% (61.7, 72.7) correctly, versus non-users' 57.6% (54.9, 60.3). distinguish between human-written somewhat better than flipping fair coin, against our initial hypothesis. Rigorously planned are needed elucidate risks benefits integrating such technologies routine practice.

Language: Английский

---

Navigating the EU AI Act: implications for regulated digital medical products

npj Digital Medicine, Journal Year: 2024, Volume and Issue: 7(1)

Published: Sept. 6, 2024

Language: Английский

---

Possible harms of artificial intelligence and the EU AI act: fundamental rights and risk

Journal of Risk Research, Journal Year: 2024, Volume and Issue: unknown, P. 1 - 14

Published: May 11, 2024

Various actors employ the notion of risk when they discuss future role Artificial Intelligence (AI) in society – sometimes as a general pointer to possible unwanted consequences underlying technologies, oriented towards political regulation AI risks. Mostly discussed within legal or ethical framework, we still lack perspective on risks based sociological research. Building systems-theoretical thinking about and society, this article analyses potential limits risk-based AI, particular with regard harm fundamental rights. Drawing Act, its earlier drafts related documents, paper how regulatory framework delineates harms which implications chosen delineation has for regulation. The results show that rights are invoked rules, values foundation trustworthiness parallel being identified at from AI. attempt frame all terms creates communicative paradoxes. It opens door classification high-risk systems well standard-setting is removed systematic concerns values. additional systemic risk, addressing general-purpose models, further reveals problems delineating In sum, Act unlikely achieve what it aims do, namely creation conditions trustworthy

Language: Английский

---

Bias in adjudication: Investigating the impact of artificial intelligence, media, financial and legal institutions in pursuit of social justice

PLoS ONE, Journal Year: 2025, Volume and Issue: 20(1), P. e0315270 - e0315270

Published: Jan. 3, 2025

The latest global progress report highlights numerous challenges in achieving justice goals, with bias artificial intelligence (AI) emerging as a significant yet underexplored issue. This paper investigates the role of AI addressing within judicial system to promote equitable social justice. Analyzing weekly data from January 1, 2019, December 31, 2023, through wavelet quantile correlation, this study examines short, medium, and long-term impacts integrating AI, media, international legal influence (ILI), financial institutions (IFI) crucial factors Sustainable Development Goal 16 (SDG-16), which focuses on findings indicate that ILI, IFI can help reduce medium long term, although their effects appear mixed less short term. Our research proposes comprehensive policy framework addresses complexities implementing these technologies system. We conclude successfully requires supportive environment embraces technological innovation, backing, robust regulation prevent potential disruptions could reinforce inequalities, perpetuate structural injustices, exacerbate human rights issues, ultimately leading more biased outcomes

Language: Английский

---

Bridging the Gap in the Adoption of Trustworthy AI in Indian Healthcare: Challenges and Opportunities

AI, Journal Year: 2025, Volume and Issue: 6(1), P. 10 - 10

Published: Jan. 13, 2025

The healthcare sector in India has experienced significant transformations owing to the advancement technology and infrastructure. Despite these transformations, there are major challenges address critical issues like insufficient infrastructure for country’s huge population, limited accessibility, shortage of skilled professionals, high-quality care. Artificial intelligence (AI)-driven solutions have potential lessen stress on India’s system; however, integrating trustworthy AI remains challenging due ethical regulatory constraints. This study aims critically review current status development systems Indian how well it satisfies legal aspects AI, as identify opportunities adoption sector. reviewed 15 articles selected from a total 1136 gathered two electronic databases, PubMed Google Scholar, project websites. makes use Preferred Reporting Items Systematic Reviews Meta-Analyses Extension Scoping (PRISMA-ScR). It finds that existing studies mostly used conventional machine learning (ML) algorithms artificial neural networks (ANNs) variety tasks, such drug discovery, disease surveillance systems, early detection diagnostic accuracy, management resources India. identifies gap various associated with it. explores developing settings, prioritizing patient safety, data privacy, compliance standards.

Language: Английский

---

Algorithms for a new season? Mapping a decade of research on the artificial intelligence-driven digital transformation of public administration

Public Management Review, Journal Year: 2025, Volume and Issue: unknown, P. 1 - 35

Published: Jan. 18, 2025

This article reviews a decade (2013–2023) of scholarly discourse to deepen our understanding the AI-driven digital transformation public administration (PA). Structural topic modelling and manual coding 169 articles that focused on contextual conditions, mechanisms, outcomes, policies revealed various topics. Findings show focus and, lesser extent, with algorithmic decision-making as key theme. Issues like trustworthiness, bias, accountability in bureaucracies are highlighted. Research gaps include insufficient exploration conditions policy implementation, narrow organizational ethical outcomes. A research agenda is suggested.

Language: Английский

---

Blockchain technology for a trustworthy social credit system: implementation and enforcement perspectives

Cluster Computing, Journal Year: 2025, Volume and Issue: 28(3)

Published: Jan. 21, 2025

Language: Английский

---