Subscribe to regular updates:


Handbook for the Mapping of AI Rules

The Digital Policy Alert's taxonomy to compare international AI rules

Report Image

The Digital Policy Alert’s detailed mapping systematises AI rules from across the world. Our taxonomy provides a common language for the characteristic elements of AI rules, including regulated entities and regulatory requirements. We release this taxonomy as a enriched with a novel policy analysis tool, enriched and an inaugural analytical series to contribute to the debate on emerging AI rules.


Johannes Fritz, Tommaso Giardini

Date Published

30 Apr 2024

A “common language” to compare diverse AI rules

"If it looks like a duck, swims like a duck, and quacks like a duck, then it probably is a duck." The value of our taxonomy is to assign a common language to similar elements named differently across legal traditions. Especially in an emerging policy area, terminology is vague and can be inconsistent. To avoid unintended digital fragmentation from unclear definitions of requirements and core concepts, the DPA AI taxonomy helps stakeholders navigate and compare related provisions from different jurisdictions. Creating our taxonomy with confidence is our first service to create a common language across different regulatory environments.

A suite of analytical tools to explore legal text

Emerging AI rules address a cutting-edge technology, but still reside in 20th century technology: the PDF (at best). To improve the accessibility of AI rules, we have developed CLaiRK, an online suite of legal text analysis tools. CLaiRK simplifies how users navigate, compare, and interact with AI rules. Navigation is enabled by our formatted text interface, which highlights the relevant passages based on users’ selection of mapping tags. Comparison is facilitated by a technology that leverages our mapping to find corresponding passages in different AI rulebooks and visually highlights differences. Finally, interaction with AI rules is powered by our Retrieval-Augmented Generation chatbot, which is trained exclusively on the legal text of AI rules and cites detailed sources with each answer. 

An analytical series to highlight similarities and differences

Our analytical series demonstrates the value of our detailed mapping. The series is designed with the two goals of supporting interoperability of emerging AI rules and providing evidence for policy alternatives in the pursuit of a shared goal. We use the OECD AI principles as our high-level reference frame. To this end, we match our 80 regulatory tools into the OECD AI Principles. Proceeding along each OECD AI Principle, we analyse how governments use various regulatory tools, where their emphasis aligns and where it diverges. To support interoperability and highlight policy alternatives, this series delivers a common factual base of unprecedented detail and structure.