Meta has been pirating pornography for years to train AI, — lawsuit says

Meta is once again at the center of a scandal — this time over tons of pirated pornography and BitTorrent.

Strike 3 Holdings, which owns several large porn sites with a monthly audience of over 25 million users, filed a lawsuit in a California federal court. According to them, Meta has been downloading and distributing their movies via BitTorrent for years, probably to train its AI models.

This echoes another case that book authors are pursuing against Meta. They accuse the company of downloading at least 81.7 terabytes of pirated content from shady online libraries. Even you can find which specific books. At the time, the corporation denied the allegations by saying that it had never used torrents. However, the current evidence from Strike 3 could change the course of the trial.

The owner of the adult website said that it discovered the massive distribution of its movies after checking its own monitoring system. According to their data, since at least 2018, IP addresses associated with Meta or its employees have downloaded more than 2,390 adult videos. Some of these IPs belonged to the company itself, while the rest were disguised — through cloud services or using private addresses.

Strike 3 believes that Meta had been sitting on these distributions for days or even weeks. And it is unlikely that the goal was just browsing. It is assumed that the company wanted to speed up the download of other data. BitTorrent has a principle: more you share — faster you download. Therefore, the corporation could have «fed» the network with adult videos to have an advantage when uploading other files for further AI training.

Such activity contradicts not only copyright but also US laws. Some states have officially introduced age verification before showing 18+ videos. And if Meta was indeed downloading and distributing these movies uncontrollably, some of them were released for free.

The lawsuit also mentions at least five cases where Meta manually chose which movies to distribute during peak download periods. The plaintiff believes that Zuckerberg’s actions do not resemble the behavior of an ordinary user: too much traffic, too regular activity, and «hand» AI that automatically processes large amounts of data.

The plaintiff also filed a complaint that Meta «could use» these videos to develop its own adult content generator. It is simply that such videos contain rare examples of» human interaction, facial expressions, bodily dynamics — which is lacking in ordinary videos and is valuable for generative models.

The company claims more than 100 thousand cases of illegal distribution from IP addresses associated with Meta. Some of them have even been registered to the home address of one of the company’s employees — possibly to hide the true source of the activity.

In the lawsuit, Strike 3 demands:

• compensation for copyright infringement;
• a court injunction against Meta for further distribution of their content;
• deleting any videos that were used for AI training or have already been included in trained models.

The plaintiff is not yet naming specific Meta employees or partners who may have been involved, citing confidentiality. The response of the accused company was obvious and short:

The owner of the adult websites hopes that their data will help the authors in a parallel lawsuit to prove that Meta systematically used pirated content — not only texts but also videos. If the accusations can be proven in court, it will seriously damage the company’s image as an «ethical» AI developer.

Source: Ars Technica