Regulatory Learning Space Framework

Overview

A three-axis framework for mapping where regulatory learning activities occur and how they should be coordinated. Developed for the EU AI Act but applicable to any complex technology governance challenge.

The Three Axes

Axis 1: AI System Types What kind of system is being learned about?

  • Prohibited practices
  • High-risk AI under Annex I (existing product legislation)
  • High-risk AI under Annex III (new risk categories)
  • General Purpose AI
  • GPAI with systemic risk
  • Transparency-obligated systems
  • Voluntary code participants

Axis 2: Protections What are we trying to protect?

  • Health
  • Safety
  • Fundamental rights
  • Democracy and rule of law
  • Environment

Axis 3: Learning Activities What type of learning is occurring?

  • Individual (literacy, competence)
  • Organizational (internal preparation)
  • Enforcement (authority interactions)
  • Meta-learning (revising rules from observed patterns)

How to Use It

Position any specific learning activity in this 3D space. This reveals:

  • Which actors need to be involved
  • What information flows are required
  • Where coordination with adjacent activities matters
  • How learnings should propagate to meta-level revision

Why It Matters

Without structured coordination, learning activities duplicate effort, miss important interactions, and fail to inform the meta-level improvements that make regulation adaptive. The framework makes the complexity of regulatory learning visible and manageable.

Limitations

The framework describes structure, not dynamics. It doesn’t specify how information should flow between positions in the space, that requires separate interoperability standards and governance agreements. It also doesn’t address how to resolve conflicts when learnings in different parts of the space suggest contradictory directions.

Related: 05-molecule—regulation-as-learning-framework, 05-atom—horizontal-vertical-regulation-tension, 05-molecule—technology-pacing-problem