crossroads magazine Luxembourg’s open-source AI Assessment Sandbox Configurator helps international companies set up a controlled assessment environment to generate evidence of the AI systems’ trustworthiness. BY EMILIO NAUD Companies looking to deploy an AI system in the Eu- ropean market face a fundamental challenge: how to test and document their solution without losing control of their data, infrastructure or evidence. Luxembourg helps address the issue with its latest AI Assessment Sandbox Configurator. Turning European requirements into evidence An AI system cannot be assessed through model perfor- mance alone. Depending on its use cases, organisations may need to assess its robustness, hidden bias, explaina- bility, privacy and regulatory compliance, to name a few. In addition, the EU AI Act requires demonstrating the assessment in detail. This means translating legal requirements into controls and metrics, se- lecting relevant tests, documenting the assessment and its results and preparing evidence for review or audit. This is often a manual, lengthy process that re- quires many multi-disciplinary profiles. Luxembourg’s AI Assessment Sandbox Configura- tor considerably simplifies this process. Released by the Luxembourg AI Factory in June 2026, it is a free, open-source tool for building customised AI testing environments. Instead of starting from scratch, us- ers first qualify their AI system, then select the tests, controls and data sets relevant to their use case from a curated catalogue and adapt them to their needs. This matters because the risk of an AI system is spe- cific to its use case. For instance, a financial service, a public-sector application and an industrial system will likely require different tests and controls. The tool then generates a controlled testing en- vironment with a collaborative dashboard, a report © BoliviaInteligente