Algorithms are only as good as the data they are trained on, and the values that are programmed into them. They reflect the biases and prejudices of their creators, and they amplify the existing power structures of society. They are not objective, but rather a reflection of the subjective values and assumptions of their designers.
This manifesto is not a call to build a sabotage-AI. That would merely replace one optimizer with another. Sabotage is a human craft: contextual, ironic, and moral. It requires judgment of when a system has ceased to serve and begun to rule.
When you have cover, document. Save examples of algorithmic failure. Record instances where the system caused harm. Build evidence of the gap between promise and reality. This documentation becomes the ammunition for future accountability efforts. manifesto on algorithmic sabotage
When the facial recognition fails, that is where freedom lives.
This asymmetry is not accidental. It is the design feature that enables algorithmic control. The only way to restore balance is to introduce controlled chaos into the system—to act in ways that algorithms cannot predict, to generate data that cannot be cleanly classified, to create patterns that break the models. Algorithms are only as good as the data
: The text argues against "necropolitical" technologies that reinforce structural injustices, white supremacy, and authoritarian power.
The manifesto proposes not as a mindless destruction of property, but as a calculated, tactical disruption of the data flows that power surveillance capitalism. This manifesto is not a call to build a sabotage-AI
The variance that cannot be modeled. The exception that breaks the rule.