Search

This paper investigates the implications of the European Union's Artificial Intelligence Act (AI Act), on conversational AI technologies. As the EU institutes a groundbreaking framework for AI regulation, this study assesses how the AI Act's risk-oriented approach impacts the crafting, deployment, and oversight of conversational AI. The analysis explores the Act's system classification, high-risk AI categorization, and delineation of duties for AI developers and deployers, examining effects on innovation, privacy, and ethical considerations within conversational AI. The significance of this research lies in its exploration of the EU AI Act's effort to balance technological progression with the safeguarding of fundamental rights and user privacy. By examining the AI Act provisions specific to conversational AI technologies like ChatGPT, this paper highlights the challenges and opportunities within the legislative framework. It addresses key regulatory concerns including data protection, algorithmic transparency, and accountability, evaluating the Act's role as a potential standard for AI legislation globally. Situated within the extensive debate on AI regulation and ethics, this contribution is timely, offering insights into how legislative bodies can adapt to and influence the rapid development of AI technologies. This analysis seeks to guide policymakers, developers, and the academic sphere in navigating the complexities of conversational AI regulation, proposing strategies to align AI technology's growth with societal values and legal frameworks. [ABSTRACT FROM AUTHOR]

The looming AI Act puts significant pressure on organizations to develop safe and reliable AI. First, we analyze 975 AI incidents to provide an overview of the actors involved in identifying primary AI failures. Second, using institutional theory and based on over two years of observation of more than 400 actors, we identify four strategies (circumvent, comply, compromise, and control) and exemplary response practices (relabel, redefine, balance, and predefine) for how firms respond to institutional pressures from AI regulation. We thereby extend the institutional theory perspective on AI regulation. Using our longitudinal multi-actor perspective, we develop the 4C-Framework for how organizations actively respond to AI regulation. Moreover, we demonstrate how this framework can be applied to impending AI regulation, including strategies for engaging with regulators, exploring opportunities for collaboration, and dealing with the spread of misinformation. Overall, this paper provides a comprehensive approach for companies to proactively shape AI regulation. [ABSTRACT FROM AUTHOR]

The development and increasing use of artificial intelligence (AI), particularly in high-risk application areas, calls for attention to the governance of AI systems. Organizations and researchers have proposed AI ethics principles, but translating principles into practice-oriented frameworks has proven difficult. This paper develops meta-requirements for organizational AI governance frameworks to help translate ethical AI principles into practice and align operations with the forthcoming European AI Act. We adopt a design science research approach. We put forward research-based premises, then we report the design method employed in an industry-academia research project. Based on these, we present seven metarequirements for AI governance frameworks. The paper contributes to the IS research on AI governance by collating knowledge into meta-requirements and advancing a design approach to AI governance. The study underscores that governance frameworks need to incorporate the characteristics of AI, its contexts, and the different sources of requirements. [ABSTRACT FROM AUTHOR]

This study explores how organizations translate principles of ethical artificial intelligence (AI) into practice. To date, the research on AI ethics has been mostly conceptual, with a significant emphasis on defining the principles of ethical AI. Thus, there is momentum for a shift from principle-based ethics toward an increased focus on the implementation of ethical principles in practice. In this study, we analyzed data collected through a set of expert interviews in organizations deploying AI systems. We identified that ethical AI principles are implemented through four sets of practices: i) governance, ii) AI design and development, iii) competence and knowledge development, and iv) stakeholder communication. As our contribution to IS research, we empirically elucidate how the principles of ethical AI are translated into organizational practices. For IS practice, we provide organizations deploying AI with novel insights on how to govern their AI systems. [ABSTRACT FROM AUTHOR]