AI projects launched in Latvia involving pre-schoolers and diabetics

To the potential concern of persons not enthusiastic about rapid AI adoption and data sharing, these first systems being launched are related to some of society's more vulnerable members: pre-schoolers, and persons with type 1 diabetes. However, the regulatory sandbox the projects are using is intended to allay such fears. “Both projects affect children’s data in the most sensitive areas, education and healthcare, and that is why they are being developed in a special regulatory environment, not outside of it. This is not an exemption from regulatory requirements, but a structured path to compliance, in which the company receives assessments from competent authorities, documented operating limits and clarity about what is permissible already at the development stage, not after the market launch. A company developing a high-risk artificial intelligence system in Latvia has not had such a path so far,” said Guna Puce, Director of the Artificial Intelligence Centre. MyWisy is an adaptive learning platform in preschool education for children aged 4 to 7, developed in Latvia, that personalizes the learning content, task sequence and pace according to each child’s individual learning level. The platform is being developed in two separate versions, for the family environment and for educational institutions, with different functionality and data processing. “When creating MyWisy, it was clear to us from the very beginning: working with preschool children's data requires not only a vision, but also a reliable technological foundation. The AI Centre sandbox provides us with both clarity regarding documentation and a very practical benefit... This practical approach is already giving us a significant breakthrough in active negotiations with foreign partners, allowing us to scale the solution globally much faster and more confidently,” said Andis Kulinskis, co-founder of SIA MyWisy. Meanwhile ANIMAS-DA is a digital artificial intelligence-driven decision support tool for planning daily insulin dosing in pediatric type 1 diabetes. The system compiles continuous glucose monitoring data, and other indicators, providing recommendations on carbohydrate intake, the safe interval for bolus correction, and the prediction of the risk of hypoglycemia or hyperglycemia. “When creating the Diabetes Assistant, one of the main tasks at the early stages of development was to determine a safe, responsible and legally justified approach to the use of real health data, including training the artificial intelligence model. Therefore, we chose a non-clinical mode as the first stage of development, which allows us to evaluate data processing, model operation and security mechanisms in a controlled environment. Timely consultations with the State Agency of Medicines and the State Data Inspectorate help to clarify the applicable requirements and properly prepare for the compliance assessment already at the development stage,” said Gatis Romanovskis, Head of SIA “Animas”. The MIC organizes a special regulatory environment for AI, coordinates cooperation between companies, state institutions and regulators, and also develops methodological guidelines for developers of AI systems. More information here.