To avoid blocking AI-driven innovations for advancing healthcare systems the use of synthetic data generators has become a priority within both research and industry. Given the success of large language models the question remains open on how to create synthetic journal data that are both realistic, able to be safely utilized in downstream AI-driven tasks and still be able to exhibit privacy-preserving properties. This lecture will present an ongoing multidisciplinary research project within Västra Götalandsregionen with the goal to highlight the potential and challenges in creating synthetic patient journals for AI development in primary care. The focus will be on legal, algorithmic and evaluation perspectives.