Toward a Human-Centred Framework for Trustworthy, Safe and Ethical Generative Artificial Intelligence

This dissertation explores the trustworthy, safe, and ethical use of Generative Artificial Intelligence (GenAI) through an examination of the human rights risks and potential social hazards these systems pose. Employing a mixed-methods approach, this study adopts a multidimensional perspective that engages various stakeholders—including the AI industry, governmental institutions, and non-profits—to analyse current risks and the strategies AI companies need for ethical development. From this analysis, two instruments are designed to navigate societal challenges: 1) the Human Rights Impact Assessment Tool (HRIAT), a structured methodology to identify and mitigate human rights risks of GenAI models, and 2) the AI Ethics Incentive Scheme (AIEIS), a governance framework that introduces reputational, financial, and collaborative incentives to encourage proactive ethical practices and cross-sectoral cooperation. These tools were developed in collaboration with the Institute for Human Rights at the University of Alabama at Birmingham and constitute an original contribution to the current debate around responsible innovation. The proposed tools aim to ensure that AI innovation advances in step with human rights principles and ethical standards, moving beyond mere regulatory compliance toward a human-centred, trustworthy, and ethical approach.