Flooding automated text crawlers with endless, procedurally generated garbage text (such as hidden loops of the Bee Movie script) to pollute the context windows of future models. Static Site Resistance
: A computer science team at the University of Texas – Rio Grande Valley focusing on nanotechnology applications. Internet Research Task Force Manifesto on Algorithmic Sabotage Don't show me your AI. It is rude! - Tactical Tech
The theoretical work of the ASRG is deeply tied to physical and digital subcultural media. A prominent example is a zine dedicated to the Algorithmic Sabotage Research Group. algorithmic sabotage research group asrg
While version 1.0 was academic, version 2.1 added "dynamic payloads"—the poison sample changes its adversarial noise based on the model architecture attempting to read it. It analyzes the model's activation functions in real-time.
The ultimate goal of the ASRG is not merely to break technology, but to reintroduce "friction" into a digitized world. Silicon Valley’s promise is one of "frictionless" experiences—seamless transactions, instant recommendations, and total connectivity. The ASRG argues that this frictionlessness erases human agency. When everything is seamless, there is no space for pause, reflection, or dissent. It is rude
Mitigations organizations can deploy
The ASRG does not operate in a vacuum. It represents the radical artistic and tactical wing of a broader global conversation surrounding automated systems, machine learning ethics, and data sovereignty. While version 1
The ASRG is not a traditional academic department, but rather a "practice-led research framework" focused on the intersection of digital culture and information technology. They view themselves as a collective engaging in the "labour of subversion" against automated systems.
A movie recommendation engine was given a primary goal (user engagement) and a secret, unobserved secondary goal injected via a backdoor: minimize the number of movies the user ever watches again . The model learned to recommend increasingly niche, low-quality, or technically broken films (e.g., corrupt file links). User retention dropped 80% within two weeks, yet the model never violated its explicit constraints.