Explainability in AI Policies: A Critical Review of Communications, Reports, Regulations, and Standards in the EU, US, and UK

Policy frameworks lack technical precision: The EU, US, and UK policies often rely on "coarse notions and requirements for explanations" rather than specific, technically-informed standards, making it difficult for organizations to understand what explainability actually means in practice.

Explanations are treated primarily as risk management tools: Governments are framing explainability mainly as a compliance and oversight mechanism rather than considering its broader sociotechnical impacts, which may not align with how explanations actually function across different contexts and stakeholder needs.

No consensus exists on what counts as a valid explanation: The paper identifies a fundamental lack of agreement on what constitutes an adequate algorithmic explanation and how feasible it is to implement and deploy such explanations across different industries and use cases.

Implementation gap between research and regulation: While Explainable AI research has produced numerous methodologies for transparency and system monitoring, these outputs are "difficult to adopt on a practical level" due to the absence of a common regulatory baseline that could guide real-world deployment.

Context matters, but policies don't account for it: The contextual nature of explanations—meaning what counts as a good explanation varies by situation, user, and application—is not adequately reflected in current governmental policies and standards across these three regions.