Poison Data, Kill Algorithms
Belgium
2026
Poison Data, Kill Algorithms explores data collection as a site of resistance through the concept of data poisoning. Private surveillance companies are constantly trying to supply corporations and governments with the most advanced monitoring tools, to the detriment of citizens’ rights. To train their algorithms, they collect enormous amounts of public data, and heavily rely on outsourced labour to create image datasets.
The latter are often designed to anticipate future flaws in surveillance. The ones built to improve “liveliness” detection contain images of people wearing different disguises in various combinations. After researching some of these datasets available online, Marcel Top created his own.
However, as 10% of Top’s dataset is poisoned, its use is likely to deteriorate any algorithm it trains, making it unreliable, therefore disrupting unethical automated surveillance and data collection.
Portfolio
About the artist
Born in 1997, Marcel Top lives and works in Belgium. He researches the topics of mass surveillance, privacy, data collection. In his practice, he layers a traditional approach to documentary research with a more experimental use of new technologies (such as facial recognition, movement analysis, and deepfakes). The artist uses these technologies to visualise and examine scenarios in which people can protect themselves and their rights by gaining knowledge and reclaiming control of surveillance tools.