Vixen Media Group has some of the sexiest content in the adult industry. So it’s no surprise that AI companies might want to use it to train their AI models, but the problem is, they didn’t get permission, so Vixen is suing.

But Meta says not so fast …. Meta Platforms, the parent company of Facebook, has asked a federal court to dismiss a copyright infringement lawsuit filed by adult entertainment company Strike 3 Holdings, which accused the tech giant of downloading thousands of pornographic films to train its artificial intelligence models.

The complaint, filed in July by Strike 3—the parent company of Vixen Media Group (VMG)—alleges that Meta used a network of roughly 2,500 hidden IP addresses to download and distribute nearly 2,400 of its copyrighted films without authorization. The company claimed that the material was used to train an “adult-friendly” version of Meta’s upcoming AI video generation tool, known internally as Movie Gen, and sought more than $350 million in damages.

In a motion to dismiss filed Monday in the U.S. District Court for the Northern District of California, Meta called the allegations “unsubstantiated” and “nonsensical,” arguing that Strike 3 failed to provide any evidence connecting the alleged downloads to Meta’s employees or AI projects.

“The complaint fails to state facts sufficient to support a claim for direct, vicarious, or contributory copyright liability,” wrote Angela L. Dunning, a partner at Cleary Gottlieb Steen & Hamilton LLP, representing Meta. “Plaintiffs’ theory of liability makes no sense and cannot be reconciled with the facts they do plead.”

Strike 3 claimed that the downloads originated from IP addresses registered to Meta and that the company concealed its activity using a “stealth network.” Meta rejected that characterization, saying that the limited torrenting activity identified by Strike 3 likely came from individuals using its vast corporate network—possibly contractors or visitors—rather than from Meta’s AI research teams.

“The far more plausible inference,” Meta’s filing argues, “is that disparate individuals downloaded adult videos for personal use.”

Meta also pointed out that the alleged downloads occurred intermittently over a seven-year period, beginning in 2018, years before the company began serious development of multimodal or generative video AI systems.

“It is implausible that a few dozen titles per year, downloaded one file at a time, could serve as training data for modern AI models,” the filing said.

Strike 3, known for filing thousands of BitTorrent-related lawsuits, frequently uses IP-tracking software to identify suspected copyright violations. Critics and digital rights advocates have long warned that attributing illegal downloads to individuals or organizations based solely on IP addresses, especially those belonging to large corporate networks, can be unreliable.

Meta further emphasized that its internal policies prohibit the inclusion of adult content in training datasets.

“We don’t want this type of content,” a Meta spokesperson said in a statement. “We take deliberate steps to prevent it from entering our AI training pipelines.”

If the court denies Meta’s motion, the case could move to discovery, where Strike 3 would seek internal network and data provenance records from Meta to substantiate its claims.

Regardless of the outcome, the case underscores growing legal scrutiny over AI training data provenance and the need for technology companies to document how datasets are collected and filtered, especially when sexually explicit or copyrighted material is involved.

Strike 3 has two weeks to respond to Meta’s motion to dismiss.