«Exposing.ai» traces image datasets collected in the wild, revealing how millions of photos uploaded to platforms like Flickr have been repurposed to train, test, and enhance facial recognition and surveillance technologies for academic, commercial, and defense applications. It offers a search engine that allows Flickr users to discover whether their images appear in face recognition datasets. The project emerged from research fellowships at the HfG Staatliche Hochschule für Gestaltung Karlsruhe Critical AI Group and the Weizenbaum Institut in Berlin, and was developed in collaboration with Jules LaPlace and the «Surveillance Technology Oversight Project» (STOP). It raises questions about the exploitation of Creative Commons licenses, the ethics of dataset collection, and the lack of transparency in AI research. The project has been featured in the Financial Times, New York Times, Nature, and a US Government Accountability Report, helping to bring the ethics of biometric data collection into mainstream public discourse.