is a form of techno-disobedience. It isn't about hating technology; it’s about subverting the harmful ways technology is used to enforce social control, labor precarity, and structural injustice.

According to their , the group rejects the idea that opposing technology is an "atavistic aversion" or a simple luddite impulse. Instead, they frame sabotage as an ethical action-oriented commitment to social autonomy and egalitarianism. Their philosophy centers on:

| Attack Surface | Target | ASRG Research Focus | |----------------|--------|----------------------| | | Labeling services (e.g., Mechanical Turk) | Subversion of annotators : paying workers to systematically mislabel a specific class (e.g., all "pedestrian" as "street sign"). | | Model Registry | Hugging Face, internal model stores | Trojan model uploads : publishing a "helpful" fine-tuned model that contains a logic bomb. | | Inference API | Public-facing ML endpoints (GPT, Claude, Gemini) | Extraction via sabotage : crafting queries that force the model into a repetitive, resource-exhaustive loop (a new form of algorithmic DoS). | | Continuous Learning Pipeline | Online retail, fraud detection | Drift injection : feeding a slow, plausible shift in input distribution so the model gradually becomes racist, sexist, or financially reckless without triggering alarms. | | Human-in-the-Loop | Content moderation systems | Overwhelming the human : generating millions of borderline-violating posts to cause moderator burnout and policy drift. |

Opposing AI and data tools used in warfare and surveillance that treat people as mere variables. Technosolutionism:

Early results, shared in a preprint, suggest that sabotage leaves a distinct in gradient updates: a kind of “stutter” in loss landscape smoothing. If validated, this could become the first practical defense against algorithmic self-sabotage.

ASRG defines its work as and practice-led . Their primary output, such as the "Manifesto on Algorithmic Sabotage," outlines 10 principles for resisting what they call "algorithmic humiliation"—the use of automated systems to maximize power and profit at the expense of human dignity. Key Themes of Their "Sabotage"