Meta is fighting back against explosive claims that it illegally downloaded thousands of adult films to train its AI models. In a motion to dismiss filed this week, the tech giant argues that any pornography found on its corporate networks was downloaded by employees for personal use, not to power an adult version of its Movie Gen AI. The lawsuit from Strike 3 Holdings could cost Meta over $350 million if successful.
Meta just delivered its most unusual legal defense yet - claiming that pornography downloaded on company networks was strictly for employees' personal entertainment, not AI training. The social media giant filed a motion to dismiss this week against Strike 3 Holdings, the adult film company that discovered illegal downloads of its content traced back to Meta's corporate IP addresses.
The case centers on Strike 3's explosive allegation that Meta secretly torrented around 2,400 adult films to train an unannounced adult version of its Movie Gen AI model. Strike 3 claims it uncovered not just downloads on traceable Meta IPs, but also evidence of a 'stealth network' using 2,500 hidden IP addresses to conceal the activity. The potential damages could exceed $350 million, according to TorrentFreak's reporting.
But Meta isn't buying Strike 3's narrative. In its filing, the company argues the downloads spanning seven years were nothing more than scattered personal use by individual employees. 'The far more plausible inference to be drawn from such meager, uncoordinated activity is that disparate individuals downloaded adult videos for personal use,' Meta's lawyers wrote.
The numbers tell Meta's story. Rather than the massive, coordinated data collection typical of AI training operations, the alleged activity amounted to roughly 22 downloads per year - what Meta calls 'a few dozen titles per year intermittently obtained one file at a time.' Compare that to the hundreds of thousands of books or images typically used in AI training datasets, and Meta's personal use defense starts looking more credible.
Timing also works in Meta's favor. The alleged downloads began in 2018, about four years before Meta's AI video research efforts even launched. 'These claims are bogus,' a Meta spokesperson told Ars Technica, emphasizing that the company's terms explicitly prohibit generating adult content - undermining any business case for training AI on such material.
The stealth network allegation particularly baffles Meta's legal team. If the company was really orchestrating a secret porn-harvesting operation, why would it simultaneously use easily traceable corporate IP addresses for 'many hundreds' of other downloads? 'The obvious answer is that it would not do so,' Meta argues, calling Strike 3's entire theory 'nonsensical and unsupported.'
Meta also plays the scale card effectively. With tens of thousands of employees, plus contractors, visitors, and third parties accessing Meta's network daily, pinning responsibility on the company becomes nearly impossible. The filing notes that Strike 3 'does not identify any of the individuals who supposedly used these Meta IP addresses' or establish whether they had any role in AI training.
One specific case involves a Meta contractor who allegedly downloaded content at his father's house. But this contractor worked as an 'automation engineer' with no apparent connection to AI training data collection. When his Meta contract ended, the torrenting allegedly stopped - but Meta argues this proves nothing about corporate direction or knowledge.
Strike 3 Holdings has built a reputation as what some call a 'copyright troll,' filing multiple lawsuits against alleged downloaders of its adult content. Meta's motion doesn't shy away from this characterization, describing Strike 3's approach as 'extortive lawsuits' based on 'guesswork and innuendo.'
For Meta, winning this case means more than avoiding a massive payout. The company is working to position its AI tools as family-friendly alternatives to competitors, with explicit content policies that ban adult material generation. Any association with training AI on pornography could damage that carefully crafted image and potentially trigger regulatory scrutiny of its AI practices.
The legal battle also highlights broader questions about corporate responsibility for employee internet use. Meta argues that monitoring every download across its global network would be 'extraordinarily complex and invasive,' and that current law only requires 'simple measures' to police such activity.
Strike 3 now has two weeks to respond to Meta's dismissal motion, setting up what could become a precedent-setting case about AI training data sources and corporate liability for employee actions.
This unusual case exposes the messy intersection of corporate AI development, employee privacy, and copyright enforcement in the digital age. Whether Meta's personal use defense holds up in court could set important precedents for how companies police their networks and what constitutes evidence of corporate-directed AI training. For now, Meta's betting that the scattered nature of the downloads and timing inconsistencies will convince the court that this was employee misconduct, not corporate strategy.
