Towards a Policy-as-a-Service Framework to Enable Compliant, Trustworthy AI and HRI Systems in the Wild

Trustworthiness is not just a technical problem: The paper argues that building trustworthy AI systems requires understanding complex human factors and relationships between agents, humans, organizations, and governments—each with their own definitions of trust. Engineers cannot simply program systems to "always do the right thing" without considering this broader socio-technical context.

AI developers lack practical tools for compliance: Systems developers generally desire trustworthy outcomes but lack practical expertise in law, regulation, policy, and ethics needed to ensure compliance. This gap between intention and capability is a significant barrier to deploying trustworthy AI systems in real-world settings.

Policy must be considered during design, not after deployment: The paper emphasizes the need for "trustworthy-by-design" interventions integrated into the systems engineering process from the start, rather than treating compliance and policy considerations as afterthoughts once systems are already built.

A "Policy-as-a-Service" framework can help bridge the gap: The authors propose a PaaS framework that AI systems engineers can readily apply to address trustworthiness challenges, suggesting that making policy expertise more accessible and usable could help developers build compliant systems more effectively.

Trust is multidimensional and context-dependent: The paper highlights that trustworthiness involves "fuzzy" socio-technical aspects that vary across different stakeholders and contexts, requiring careful consideration during both the design and deployment phases of AI systems.