Meta sued by adult site for using pornographic videos to train AI, but claims it was for 'personal use'

In July 2025, Meta was sued by an adult film production company for illegally downloading approximately 2,396 adult videos and using them without permission to train its AI models. Meta responded by arguing that the downloads were for the private and personal use of its employees and site users, not for the purpose of training its AI, and that it did not use adult content to train its AI.

Meta: Pirated Adult Film Downloads Were For 'Personal Use,' Not AI Training * TorrentFreak

Adult film production companies Strike 3 Holdings and Counterlife Media filed a lawsuit in the U.S. District Court for the Northern California District on July 23, 2025, alleging that Meta downloaded at least 2,396 copyrighted videos via the BitTorrent protocol since 2018. Strike 3 Holdings is seeking up to $359 million in damages, citing allegations that Meta attempted to use the illegally obtained content in its AI-powered adult video generation tool.

Strike 3 Holdings filed 2,788 lawsuits in 2022 for alleged piracy of its content. This accounts for the majority of copyright infringement lawsuits filed in the United States in 2022 and is a record number for a rights holder to file in a single year. The practice of enforcing copyrights, such as Strike 3 Holdings', to 'track down individual uploaders of pirated content and seek settlements to gain financial benefits' has been criticized as a profit-making activity known as ' copyright trolling ,' with some claiming that ' tracking addresses based on IP addresses and demanding payment amounts to blackmail .'

Adult content companies are facing a record number of copyright lawsuits, and what are 'copyright trolls' who abuse copyrights for commercial purposes? - GIGAZINE

Meta's use of the BitTorrent protocol has been identified in past lawsuits. Strike 3 Holdings and Counterlife Media, suspecting copyright infringement, searched for Meta-linked IP addresses in BitTorrent data archives. They identified 47 IP addresses as belonging to Meta. Strike 3 Holdings claims that Meta illegally downloaded adult content from these IP addresses and used them to train commercial AI. They also claim that Meta may have used approximately 2,500 hidden IP addresses to conceal its illegal downloads.

Meta countered that Strike 3 Holdings' claims were based solely on IP addresses and could not identify who actually downloaded the content and what they used it for, making them insufficient to assert a direct claim of copyright infringement. According to Meta, the copyright infringement records date back to 2018, but Meta only began researching multimodal video generation models in 2022, contradicting the timing of the claim that Strike 3 Holdings 'obtained content to train its AI models.'

Furthermore, Meta argued that the recorded downloads, averaging around 22 per year from several dozen IP addresses, 'clearly represent personal use, not a concerted effort to collect the massive datasets Plaintiffs claim are necessary for AI training.' While Meta does not deny the possibility that its network was used to download pirated adult content, it argues, citing precedents from previous BitTorrent-related litigation, that Meta cannot be held secondarily liable because it did not financially benefit from these 'personal use' downloads.

In response to the allegation that the plaintiffs used approximately 2,500 secret IP addresses to conceal illegal downloads, Meta responded, 'Why conceal certain downloads while using easily traceable Meta corporate IP addresses for hundreds of downloads, including 157 of the plaintiffs' works?' It also stated, 'Plaintiffs' entire AI training theory is nonsensical and unsupported.'

Meta has filed a motion in response to the lawsuit, giving Strike 3 Holdings two weeks to respond. After Meta files a follow-up response, a California federal court is expected to decide whether to continue the lawsuit.