A growing danger is hidden in the volumes of video and images accumulated online. Using this large-scale pixel data, artificial intelligence algorithms can discover privacy-sensitive information about people, such as: location, personal possessions, socio-economic status, health state, psychological profile, and daily patterns. Currently, the sheer scale of data stored, and the frequency of data breaches makes visual privacy more important than ever. Trusting today’s data-addicted companies, such a Facebook, Google, Twitter and Instagram, to solve the problem is effectively leaving the fox to guard the chicken coop. This project develops “Pixel Privacy”: it provides users with visual privacy protection for their images and videos, freeing them from dependence on external parties. The project takes the idea of a multimedia search engine and inverts it. Rather than using the search engine to find information, it uses the search engine to determine what must be kept hidden. Efficient big-data algorithms identify visual elements related to implicit personal information, and substitute safer elements. This fundamentally new idea elegantly avoids the disadvantages of mask-everything approaches and cryptography-based systems. Pixel Privacy guards against criminal attacks (such as targeted home robbery), but, crucially, also safeguards the autonomy of citizens, which is essential for healthy democracy.