Reed Ghazala’s concept of circuit bending transforms electronic devices into “living instruments” with new, unpredictable behaviours. This workshop extends that philosophy into network bending, applying it to generative neural networks for sound design, live performance, and experimental sound art. Participants will explore intuitive manipulation of open-source neural networks without needing deep technical expertise. Like circuit bending, network bending embraces unpredictability and creative disruption, encouraging participants to subvert conventional design and sound aesthetics. The workshop emphasizes accessibility by using small, interpretable, and computationally efficient networks trained on personal sound recordings. Participants will gain hands-on experience with network bending techniques, training neural networks online or locally, and experimenting with tools like Pure Data to produce unconventional sounds. This approach fosters an open, anti- design methodology, allowing for experimentation beyond traditional boundaries and challenging familiar aesthetics in sound creation.