Meta Accused of Torrenting Porn to Advance Its Goal of AI ‘Superintelligence’

Summary

Strike 3 Holdings has sued Meta in federal court, alleging the company pirated thousands of copyrighted adult videos and other copyrighted content via the BitTorrent protocol to train its AI models. The unsealed complaint claims Meta torrent-seeded 2,396 Strike 3 videos since 2018 to obtain rare visual data — extended, uninterrupted scenes and unusual camera angles — that could improve the realism and “humanity” of its models. The suit alleges this made content accessible to minors and that Meta used porn as a form of “currency” to access more data. Meta says it is reviewing the complaint and disputes the claims. Strike 3 is seeking $350 million in statutory damages. The case is part of a broader wave of AI copyright litigation testing fair use and training-data practices.

Key Points

Strike 3 alleges Meta used BitTorrent to download and seed thousands of copyrighted porn videos (2,396 titles listed) beginning in 2018.
The complaint asserts Meta sought adult content for specific visual features — long takes, body angles and continuity — valuable for training generative video models.
Exhibits also list mainstream TV shows and other copyrighted material; some non-Strike 3 titles raise concerns about underage performers.
Strike 3 claims Meta distributed the files (via BitTorrent), which lacks age verification, potentially exposing minors to adult content.
Meta’s V-JEPA 2 research says it trained on “one million hours of internet video” — the complaint argues that term is unspecified and may include pirated content.
Strike 3 is seeking $350 million; the case follows other major copyright suits and could shape how AI companies gather and justify training data.
Legal precedent is unsettled: a separate ruling in Kadrey v. Meta sided with Meta on pleading issues but did not establish a broad legality for training on copyrighted works.

Context and Relevance

This lawsuit lands at the intersection of AI development, copyright law and content safety. As large tech firms race to build ever-more capable models and “personal superintelligence,” the sources of training data are under intense scrutiny. Courts are being asked to balance claims of market harm and unauthorised copying against tech companies’ arguments that model training is transformative and protected as fair use. High-profile cases (and large settlements elsewhere) show the financial and reputational risks for firms found to have used copyrighted material without permission. For regulators, creators and platform operators, the outcome could influence data-acquisition practices, age-safety obligations and the broader economics of content licensing for AI.

Why should I read this?

Short version: this is messy, important and could change how AI firms get their training data. If you care about who gets paid for creative work, how safe AI outputs are for kids, or whether big tech can quietly raid the internet to build “superintelligence”, read this. It’s the kind of case that might set new rules — or at least make companies think twice before downloading someone else’s content.

Source

Source: https://www.wired.com/story/meta-lawsuit-strike-3-porn-copyright-ai/