Violations of copyright, private data and fake news: for Dutch texts, OpenAI’s filter is of rather poor quality, conclude Joris Veerbeek (Data School) and Eva Hofman (De Groene Amsterdammer). In this article they investigate which Dutch sources are used to train chatbots such as ChatGPT. Read article (in dutch)

The Utrecht Data School (UDS) regularly collaborates with De Groene Amsterdammer on journalistic articles that require thorough data research. Find all articles here.