MARLA: A Conceptual Scaffold for Regulatory Learning under the EU AI Act

To achieve the goal of establishing regulatory frameworks for safe, trustworthy, and market-ready innovative infrastructure in accordance with the EU’s Artificial Intelligence Act, researchers proposed the MARLA (Map, Assess, Report, Learn, Adapt) conceptual framework. This framework aims to create a five-stage cycle that transforms evidence generated by law enforcement requirements into governance and legal knowledge that supports consistent interpretation, effective regulation, and technological adaptation. The MARLA framework is intentionally non-mandatory, providing shared terminology for technical and legal stakeholders and encouraging the first three stages to produce documented forms of regulatory learning. The study illustrates the application and practical implementation of this framework through two pilot cases and a prospective case ranging from the national level to the EU level.