Dataharvest 2025 - the European Investigative Journalism Conference

Large language models and other AI systems sometimes only work because there are thousands of people labeling and classifying information that is then used in training the models. Journalistic investigations are discovering rights violations and misdemeanors by large outsourcing companies, who offer jobs below the minimum wage. In some cases, workers are forced to go through datasets of disturbing data that can affect their mental health. However, tracing this labour chain is not easy. Big Tech companies often turn to underdeveloped countries where they can take advantage of economic conditions and labour needs to employ this activity; scammers take advantage of this under-supervised process. How can you track this outsourcing chain when you are sitting at your desk somewhere in Europe? Journalists who attend this session will learn how to investigate large tech companies and follow the traces of their activities, as well as uncovering labour and human rights violations along the AI supply chain. They will also learn how to contact the right sources and approach workers who are in sensitive situations.