Google researchers have released details of a machine learning system that took six months to terminate 50,000 clusters of 130,000 channels. Rather than reviewing suspicious videos one by one, the system targets a coordinated, AI-generated spam network.
The system, called the Scalable Cluster Termination System (S-CTS), is described in the paper. Scalable detection of adversarial synthetic slop and tailored media exploitation: A LoRA-enabled multimodal defense system. Researchers say this was introduced on a “major online video platform”.
However, Google has not confirmed that the unnamed platform is YouTube.
This paper was highlighted by editor and CEO Jim Louderback. Inside the creator economyin his newsletter, linked the research to YouTube based on Google’s author and platform-specific signals. This relationship is not confirmed by Google and remains an interpretation.
The study comes as YouTube ramps up its focus on low-quality AI content. In January 2026, YouTube CEO Neil Mohan said the company was “aggressively building out established systems” to combat spam and clickbait. Separately, 16 high-rating channels were reportedly removed or had their content erased. In total, it had approximately 35 million subscribers and 4.7 billion lifetime views.
