Abstract:
This paper introduces the Blockchain-Based Trustworthy Artificial Intelligence (AI) Supported by Stakeholders-In-The-Loop Model (BCTrustAI.SL) that incorporates sociotechnical components to ensure legal compliance and conformity with the emerging trustworthiness standards for AI systems. BCTrustAI.SL combines features of the Blockchain Framework for Trustworthy AI (BF.TAI), the BlockIoTIntelligence architectural model and the Society-In-The-Loop (SITL) framework, to solve the intangibility problem of the foundational normative notions of robustness, ethicality and lawfulness, and human-centredness, and help them become applicable in practice. The work contributes to the specification and operationalisation of trustworthiness as an AI attribute, highlighting the importance of a precise understanding of its institutional foundations, high order principles and concrete key requirements (data protection, data governance, technical robustness and safety, transparency, accountability, diversity and non-discrimination, and human agency and oversight) deriving from them. On a practical/technical level, BCTrustAI.SL showcases the strengths of combining Blockchain (BC) and AI to address their individual limitations, laying the groundwork for future advancements and practical applications.