A new lawsuit against Meta puts a sharp question at the center of the AI copyright debate: what, exactly, are companies allowed to use when building models that are meant to understand the world through images and video?
Strike 3 Holdings, which describes its adult videos as “high quality,” “feminist,” and “ethical,” alleges that Meta infringed its copyrighted content and used it to train AI models. The complaint was filed in July in a federal court in California, and associated exhibits and details of the complaint were unsealed last week.
The core allegation
Strike 3 says Meta has been torrenting and seeding its videos since 2018. The company claims Meta used BitTorrent to download and distribute 2,396 of Strike 3’s copyright-protected porn videos.
In plain terms, the lawsuit focuses on both acquisition and distribution. BitTorrent can be used to move large files across the internet, but the complaint argues that doing so with copyrighted material is illegal when the downloader does not have permission.
Strike 3 also alleges that the process made its porn videos accessible to minors because BitTorrent does not have age verification. The company further claims Meta used and continues to use Strike 3’s porn “for distribution as currency to support its downloading of a vast array of other content necessary to train its AI models.”
The complaint says Strike 3 identified alleged violations through infringement-detection systems it has “developed, owns, and operates.” According to the lawsuit, the alleged activity involved 47 distinct Meta-affiliated IP addresses.
Based on statutory infringement penalties, Strike 3 is demanding $350 million.
Why adult video matters to the AI claim
The lawsuit does not frame the adult content as incidental. Strike 3 argues that Meta had a reason to seek it out for AI work.
According to the complaint, the videos allegedly offered visual material that is difficult to scrape from mainstream movies and TV. Strike 3 points to visual angles, parts of the human body, and extended, uninterrupted scenes as examples of what it says Meta wanted for its AI systems.
The company connects that alleged motive to Mark Zuckerberg’s stated goal of AI “superintelligence.” Christian Waugh, an attorney for Strike 3, alleges: “They have an interest in getting our content because it can give them a competitive advantage for the quality, fluidity, and humanity of the AI.”
Matthew Sag, professor of law in artificial intelligence, machine learning, and data science at Emory University, described the use of adult content as training data as “a public relations disaster waiting to happen.” His example was a middle school student asking a Meta AI model for a video about pizza delivery and unexpectedly receiving porn.
The exhibits go beyond Strike 3
The unsealed materials described in the source article do not stop with Strike 3’s own titles. The Exhibits list also includes titles Meta allegedly took from non-pornographic sources, including episodes of Yellowstone, Modern Family, The Bachelor, South Park, and Downton Abbey.
It also lists titles of non–Strike 3 porn videos that may include very young actors, including ExploitedTeens, Anal Teens, Asian Teen Masturbation, CasualTeenSex, and EuroTeenErotica.
The list contains other categories as well. Titles relating to weapons include 3D Gun Print and Gun Digest Shooter’s Guide to the AR-15. Other listed materials include Antifa’s Radical Plan and Intellectual Property Rights in Cyberspace.
Waugh argues that the public materials do not show the full extent of what Strike 3 claims to have found. He says Strike 3 has access to a wide selection of copyrighted content that Meta employees allegedly BitTorrented, and that the unsealed exhibits are only “a thin slice of the pie.”
Meta’s response and the AI copyright backdrop
Christopher Sgro, a Meta spokesperson, told WIRED: “We’re reviewing the complaint, but we don’t believe Strike’s claims are accurate.”
The lawsuit arrives as Meta continues to describe expansive AI ambitions. According to Meta researchers, V-JEPA 2, its “world model,” was released in June and trained on 1 million hours of “internet video.” Strike 3’s complaint points out that the phrase is left unspecified.
Zuckerberg has publicly said Meta plans to put “the power of superintelligence into people’s hands to direct it toward what they value in their own lives.” He also reiterated at the Meta Connect event on Wednesday that Meta’s signature smart glasses are designed to give users “personal superintelligence.”
The case also sits within a larger dispute over copyrighted material and AI training. The source article states that using pirated material as training data was reportedly an active choice made by Meta executives and approved by Zuckerberg directly. It also notes that nearly every major AI company has been accused of similar practices in ongoing AI copyright lawsuits.
AI companies have previously argued that their technologies are transformative because they do not directly copy rights holders’ works, a position they say should support fair use. President Trump has weighed in with apparent support for tech companies, saying in July: “You can’t be expected to have a successful AI program when every single article, book, or anything else that you’ve read or studied you’re supposed to pay for.”
Why this case could matter
A separate case, Kadrey v. Meta, shows why the legal path remains open. In June, US District Court judge Vince Chhabria ruled that Meta did not violate the law when training its AI models on 13 authors’ copyrighted books in that suit.
But Chhabria also made clear that the ruling did not establish Meta’s broader use of copyrighted materials for AI training as lawful. He said the ruling “stands only for the proposition that these plaintiffs made the wrong arguments and failed to develop a record in support of the right one.”
That distinction matters for Strike 3. Sag says the strongest version of Strike 3’s argument is that going to pirate websites undermines the market for access.
Waugh frames the case as a broader rights-holder issue, not one limited to adult entertainment. “It doesn’t matter if it’s a four-sentence poem or adult entertainment," he says. "There is no appetite in this country for what AI companies appear to be doing, which is making money off the backs of rights holders who never gave permission for it.”