Verffentlichung von (automatisch) anonymisierten Gerichtsentscheidungen

Court decisions must be published to enable information, transparency, and control. Information technology not only allows for complete publication but also requires decisions on a very large scale as training data for AI processes. Nevertheless, there is a lack of freely available court decisions, as only an average of 2.3% of decisions are published. However, comprehensive anonymization is necessary beforehand. Legal practice does not sufficiently comply with this requirement due to predominantly manual anonymization. The result is not only a publication deficit but also inadequate anonymization quality, as demonstrated by a de-anonymization experiment. This leads to conclusions regarding the legal concept of anonymity and suggests improvements. To enhance anonymization quality and throughput, methods of automatic anonymization using AI are explained.