this project confronts the rising fluency of large language models, where language’s ambiguity risks becoming fully transparent to machines. through experimental dialogues with ai like chat-gbt, a curated library of confusion prompts emerges; inputs that appear coherent to humans but destabilize machine understanding. this data is transformed into a dual toolkit: a collection of ready-made confusion prompts and an evolving guide for generating new ones. published on an open-access platform, the project invites participation, making the artwork a living, collective act of resistance while from rejecting technology, it intervenes in artificial intelligence’s drive for linguistic extraction, asserting opacity as a form of human agency. when clarity enables surveillance, confusion becomes necessary defense.