Meta downloaded 2,400 porno movies for AI training

Meta has filed a motion to dismiss a lawsuit filed by Strike 3 Holdings and Counterlife Media, which is seeking $359 million in damages. The plaintiffs accuse the company of illegally uploading over 2,300 adult videos and claim that this content could have been used to train artificial intelligence, according to TorrentFreak.

Suing Meta

Strike 3 Holdings and Counterlife Media, known by the brands Vixen, Tushy, Blacked, and Deeper, claim to have discovered IP addresses belonging to Meta using BitTorrent tools. They believe these addresses were used to download material that was later used to create an “adult” version of the SI model that became the basis for Movie Gen. The plaintiffs also allege the existence of a “hidden network” of approximately 2,500 IP addresses associated with the company.

How Meta Rebuts This

In its defense, Meta claims that the first downloads date back to 2018 – several years before the company began developing large-scale language models and AI-powered generative tools, which it only began developing in 2022. The corporation emphasizes that its internal policy strictly prohibits the creation or use of adult content. According to Meta representatives, isolated downloads may be related to the personal actions of employees or third parties with access to the corporate network, rather than to model training.

In these materials, the defense also cites the example of a contractor who, according to the company, downloaded such content at home for personal reasons unrelated to his professional activities. Meta believes that the claims are unsupported by evidence and contradict the factual circumstances.

AI Training Context

The case comes amid a series of lawsuits against tech companies accused of using pirated materials to train AI systems. In August, Anthropic agreed to pay $1.5 billion to the authors, and in September, Apple became a defendant in a similar case, accused of using scientific papers to train AI models.