Key points are not available for this paper at this time.
Regulatory sandboxes have emerged as a strategic tool for testing AI in real market conditions while granting temporary regulatory flexibilities. However, existing sandboxes often expose significant gaps regarding the protection of fundamental rights. This article examines national regulatory sandboxes across multiple jurisdictions, including the EU, the UK, Russia, and Singapore, using a structured, rights-based analytical framework that focuses on four critical dimensions: privacy and data protection, safety and life, consumer protection, and the right to a healthy environment. Using comparative legal and content analysis methods, we identify recurring vulnerabilities and regulatory gaps that may increase the risk of rights violations during experimentation. The study proposes a human-rights-oriented, trustworthy AI sandbox model that ensures that AI innovation adheres to key principles of trustworthy AI. It is designed to promote responsible AI innovation by facilitating interaction and learning between regulators and AI providers while safeguarding public interest, health, safety, and fundamental rights throughout the AI system lifecycle. The model is structured around four pillars: comprehensive project risk assessment, financial accountability mechanisms (liquidity tests), reliable compensation schemes, and efficient dispute resolution through ADR and ODR. By integrating human rights, trustworthiness, and sustainability considerations, this research provides regulators and policymakers with a normative blueprint to design more responsible and resilient sandbox frameworks. The findings contribute to ongoing debates on agile governance, trustworthy AI regulation, and technology-driven sustainable development.
Gromova et al. (Tue,) studied this question.