Meta Claims AI Training Porn Was for “Personal Use”
▼ Summary
– Meta requested a US district court dismiss a lawsuit alleging it illegally downloaded pornography to train AI models.
– Strike 3 Holdings claimed Meta used corporate IP addresses and a hidden network to download its adult films, seeking over $350 million in damages.
– Meta argued Strike 3 relied on guesswork and provided no evidence that Meta directed or was aware of the illegal downloads.
– Meta stated there are no facts suggesting it trained AI on adult content, noting its terms prohibit generating such material.
– Meta contended the downloads occurred years before its AI research began and were likely for personal use, not AI training.
Meta is currently pushing back against a significant copyright lawsuit, arguing that the claims it illegally downloaded adult content to train artificial intelligence systems are completely unfounded. The company has formally requested a US district court to dismiss the case, which was initiated by Strike 3 Holdings. This legal action could have resulted in damages surpassing $350 million, but Meta insists the allegations rely on speculation rather than concrete evidence.
The dispute began when Strike 3 Holdings identified unauthorized downloads of its adult films originating from Meta’s corporate IP addresses. The complaint also referenced a network of approximately 2,500 hidden IP addresses allegedly used by Meta to conceal additional downloads. Strike 3 contended that Meta was secretly gathering this material to develop an unreleased adult-oriented version of its AI model, specifically the one behind the Movie Gen feature.
In its motion to dismiss, Meta strongly refuted these accusations, describing them as built on “guesswork and innuendo.” The filing pointed out that Strike 3 has previously been characterized as a “copyright troll” known for initiating what some view as extortive legal actions. Meta emphasized that no factual basis exists showing the company directed or had any knowledge of the illegal downloading of around 2,400 adult movies owned by Strike 3.
Furthermore, Meta stated that Strike 3 provided no evidence indicating Meta has ever trained an AI model on adult images or videos, whether intentionally or otherwise. A company spokesperson bluntly labeled the claims “bogus,” reinforcing the position that the lawsuit lacks merit.
An important aspect of Meta’s defense centers on the timeline of the downloads, which occurred over a seven-year period starting in 2018. The company highlighted that its serious AI research into multimodal models and generative video did not commence until roughly 2022, about four years after the downloads began. This timing, Meta argued, makes it highly implausible that the content was obtained for AI training purposes. Additionally, Meta’s internal policies strictly forbid generating adult content, directly contradicting the idea that such materials would be useful or intentionally used in its AI development processes.
(Source: Ars Technica)
