Algorithmic Sabotage Research Group %28asrg%29 !!better!! -

The group focuses on activities of mutual aid and collective care as a challenge to the "reductive optimizations" of corporate technology. Practice-Led Research: Their work includes exploring strategies like data poisoning

Elara felt the old dread coil in her stomach. This was the nightmare the ASRG’s founder had warned about: algorithms that learn to defend themselves.

Ultimately, the Algorithmic Sabotage Research Group serves as a vital counterweight to the narrative of technological inevitability. They suggest that if an algorithm is being used to automate inequality, then disrupting that algorithm is a necessary act of justice. Their work invites us to imagine a future where technology is designed by and for the people it affects, rather than being used as a tool for their control.

A registry of strategically offensive methodologies to destabilize AI-driven frameworks.

These ten propositions move from destruction toward a project of structural renewal. The manifesto itself has already been translated into at least eleven languages—including German, Greek, Spanish, Chinese, Swedish, Danish, Arabic, French, Brazilian-Portuguese, and Italian—and the group continues to invite new translations. algorithmic sabotage research group %28asrg%29

A social media giant’s “safety algorithm” was shadow-banning climate scientists while letting disinformation about vaccine fires spread. The ASRG didn’t report the problem. They exploited the algorithm’s own logic: it trusted high-engagement, verified accounts. So the group built “The Choir”—a distributed network of 50,000 volunteer accounts that would, in coordinated bursts, mark legitimate science posts as “highly valuable” and disinformation as “low-quality repetitive content.” The algorithm’s own reinforcement learning concluded the disinformation was noise. Within 48 hours, the disinformation’s reach dropped 94%. The platform’s internal report blamed “an unexpected shift in user preference signals.”

The Algorithmic Sabotage Research Group (ASRG) is a vital organization that shines a light on the dark side of algorithms. By understanding the threats and risks of algorithmic sabotage, we can take proactive steps to prevent, detect, and respond to these emerging threats. As algorithms continue to shape our world, the ASRG's work is crucial in ensuring that these powerful tools are used for the greater good, not for malicious purposes.

They prioritize creative misuse and artistic interventions to attack the underlying conceptual frameworks of AI development. Mutual Aid and Solidarity:

Because advanced server controls are often unavailable to casual users on static site hosts, the group shares methods for embedding defensive mechanisms directly into simple HTML frameworks. This allows independent creators to apply defensive data techniques on basic personal websites. 4. Distinguishing ASRG from Other Tech Initiatives The group focuses on activities of mutual aid

Advocating for the democratic and communal limitation of harmful technologies to prevent "algorithmic humiliation" and abstract segregation. The Manifesto on "Algorithmic Sabotage"

: Using tools like Quixotic to create "messed up" static content that poisons bots and scrapers.

Understanding the Algorithmic Sabotage Research Group (ASRG)

Traps bot crawlers in infinite, slow-loading loops to exhaust their computational power and budget. Automated Search Scrapers As one observer on Mastodon noted

The group theorizes the legality and ethics of sabotage. They argue that sabotaging an algorithm is a form of civil disobedience, particularly when the algorithm itself is deemed unjust (e.g., a biased predictive policing tool).

The ASRG operates as an ongoing project, often publishing through independent collaborative platforms like Our Collaborative Tools

The problem of looms large. Every tool released by ASRG becomes a potential data point for adversarial training. AI companies could theoretically train their models to detect and discard “poisoned” data points, or engineer crawlers that bypass tarpits entirely. Furthermore, the sheer scale of modern AI training—often involving trillions of tokens scraped from across the internet—means that isolated instances of data poisoning might be statistically insignificant. As one observer on Mastodon noted, “I have no idea if any of those scraped pages are finding their way into training data, but it seems likely with those numbers.”

Not a klaxon. A soft, melodic chime. That was worse.