The Innovation Sandbox for Artificial Intelligence (AI) gives various organisations, including start-ups, SMEs, large companies and research institutions, access to regulatory know-how and new data sources. In return, all findings and results are shared publicly.
Different from many approaches adopted in other countries, the Innovation Sandbox for Artificial Intelligence (AI) goes a step further as the selected projects are not just being reviewed, but also being put into practice.
The goal is to promote responsible innovation based on legal and ethical criteria and to support the spread of AI in public administration, business and research.
Current Projects
Organisations specialised in AI had the opportunity to submit specific projects between March and June 2022. A total of 21 AI projects were submitted. The portfolio of project proposals reflects the high relevance of AI as a cross-cutting technology that transforms all sectors. The projects submitted were evaluated based on consistent and transparent criteria. Five projects were selected and are currently being implemented.
Regulating autonomous rsystems: guidelines for manufacturers
A growing number of autonomous systems such as self-driving tractors or lawn mowers will potentially be used in public spaces. In many cases, the general regulatory environment in areas such as standardization, admission to public roads or data protection are not clear to manufacturers. This project is focused on developing comprehensive guidelines that provide manufacturers with an overview of the most relevant factors.
Infrastructure maintenance using drones: technological feasibility of runway inspections
Maintenance of infrastructure, such as bridges, roads or pylons, is often carried out by human staff. Many of the tasks involved are time-consuming and, in some cases, dangerous. This project is centred on creating high-resolution drone images to detect cracks and damage on airport runways.
Automated corrections: legal bases for AI apps in education
Education is increasingly supported by AI, including at elementary level. The basic legal parameters , e.g. in the domain of data protection, are to some extent unclear to providers, teachers, parents and pupils. This project focuses on clarifying the legal bases for use of image recognition for the automated correction of handwritten schoolwork .
Smart parking in cities: best practices for privacy by design
Parking management is a highly significant challenge for many cities and municipalities. Relevant data on utilization and occupancy of parking spaces is often lacking, which hinders informed decision-making on traffic planning or the introduction of parking guidance systems. AI-based image recognition can create the necessary data basis for this. In this project, privacy-by-design measures are being developed for the responsible use of camera systems in public spaces.
Machine translations: recommendations for public administration
Many public administration offices deal with multiple languages in their daily work. The potential for machine translations is significant. As their work often involves sensitive information and use of specific technical terminology, the question arises as to whether generic and publicly accessible translation tools may and should be used. This project focuses oncoming up with recommendations for the use of machine translation. The case studies stem from the Commercial Register Office and the Integration Office.
Outlook
The current pilot phase of the Innovation Sandbox for Artificial Intelligence (AI) will run until April 2024. The knowledge and insights gained from all of the selected AI projects will be made publicly available until then. The results will contribute to building competencies and facilitating knowledge transfer within public administration. The aim is to promote the innovation capabilities of AI organisations by clarifying regulatory questions and tapping into new data sources. Furthermore, suggestions for areas of action will be made if regulatory gaps are found within the specific projects.
