Researchers have developed an AI-powered tool to automatically detect pirated content on Telegram, which helped remove 524 pirated channels and 71 bots in 61 days.
Telegram, a messaging app developed by a Russian entrepreneur, is used as a major messaging service in Russia and Ukraine, but it is also used as a platform for sharing pirated content such as movies and TV shows. Researchers at Louisiana State University and the University of Texas at Arlington have published their analysis of Telegram, which they describe as 'the first large-scale study on video piracy on Telegram.'
[2605.08418] Binge, Bot, Repeat: Unpacking the Ecosystem of Video Piracy on Telegram
Researchers Hunt Telegram Pirates with AI Tool, Flag Hundreds of Channels * TorrentFreak
https://torrentfreak.com/researchers-hunt-telegram-pirates-with-ai-tool-flag-hundreds-of-channels/
According to researchers, Telegram has emerged as a major platform for large-scale video piracy, where copyrighted content is rapidly spread among users. However, despite its notoriety, the structural and operational dynamics of the Telegram ecosystem are not well understood. Therefore, the research team analyzed 1,057 channels that made approximately 209,000 unique posts between December 2023 and January 2026 using a 'mixed methods' approach that integrates quantitative and qualitative data.
Researchers used locally operated large-scale language models (LLMs) to label posts with AI in order to map the ecosystem. This identified 19,033 unique pirated titles across various Telegram channels, consisting of 14,632 movies and 4,401 TV shows produced by 3,941 companies. The pirated posts were found on 983 channels and had a combined total of over 4.85 billion views.
The following graph shows the top 10 companies whose content was most infringed. Toei, which owns the rights to 'ONE PIECE' and 'Dragon Ball,' had the most pirated content circulating, accounting for 17% of all titles. Netflix came in second with 15%, and Warner Bros. was third with 12.4%.
Below are the top 10 countries of origin for pirated content distributed on Telegram channels. The United States is in first place with 28.5%, followed by Japan in second place with 11.8%. The countries following in third place and beyond are the United Kingdom, South Korea, France, India, and China.
A notable finding of this study is that AI analysis techniques revealed a complex ecosystem of pirated channels on Telegram. These pirated communities on Telegram are not simply single channels sharing videos; they are intentionally designed to be resistant to takedown requests. Pirated channels employ diverse distribution methods, with content distributed across interconnected channels, bots, and backup accounts. Researchers state that approximately 94% of the channels mapped by the AI are connected to at least one other channel, many of which could not be found using traditional search methods.
One reason Telegram is used for piracy is that it supports file uploads of up to 2GB per post. This capacity is sufficient to host multiple pirated videos directly on the platform, eliminating the need to use external file-sharing services which are more susceptible to monitoring and regulation, thus offering an advantage over traditional piracy sites. In fact, this analysis identified 579 posts that directly uploaded files to Telegram.
The research team also identified numerous mechanisms that redirect users to other Telegram channels, in addition to direct file sharing mechanisms. Because more visible channels are subject to stricter monitoring and regulatory pressure, they avoid account restrictions by simply providing links to other channels that share files, rather than sharing files directly. In addition, 128 cases of multi-level redirection were identified, and 37 channels specializing in sharing a single title were also found.
The researchers estimate that the economic losses incurred by content rights holders in the posts analyzed amount to approximately $17.49 billion (approximately 2.79 trillion yen). The research team compiled their findings into a report and sent it to Telegram's anti-fraud department and 17 major rights holders in the United States.
Based on these analysis results, the researchers have also developed 'Anti-RIP,' an AI-powered tool that can detect video copyright infringement on Telegram in real time. From February 3 to April 10, 2026, Anti-RIP investigated 249,133 new Telegram channels, detecting 802 pirated channels, 299 related channels, and 108 bots. Based on reports from Anti-RIP, 524 channels became inaccessible within two weeks, and Telegram has also reportedly removed many of the individual posts that were reported.
It's important to note that the researchers acknowledge that false positives can occur with Anti-RIP. A random sample of 1,000 posts used for testing revealed that four legitimate posts were incorrectly identified as copyright infringements. While the test reported a 98% accuracy rate, the verified error rate for flagged channels during actual operation has not been made public.
The research team has released the Anti-RIP dataset as open source on GitHub.
GitHub - Scalable-Security-Research-Lab/BingeBotRepeat: Repository for the paper: Binge, Bot, Repeat: Unpacking the Ecosystem of Video Piracy on Telegram · GitHub
https://github.com/Scalable-Security-Research-Lab/BingeBotRepeat
Related Posts:
in Web Service, Posted by log1e_dh