Meta has denied reports that it downloaded 2,400 adult videos to train its AI models. The tech giant has claimed that the files were accessed for “personal use.” The company emphasized that no adult content is used in its training datasets, calling the accusations “misleading and exaggerated.
Strike 3 Holdings and Counterlife Media, the makers of popular adult entertainment brands such as Vixen, Tushy, Blacked, and Deeper, alleged that Meta downloaded at least 2,369 of their movies since 2018 to train its AI, according to TorrentFreak reports.
The company checked its BitTorrent-tracking tools designed to detect copyright infringement of its videos. This revealed several corporate IP addresses identified as owned by Meta. It has further been alleged that “Meta concealed a stealth network of 2,500 hidden IP addresses."
The lawsuit claimed that Meta secretly trained an unannounced adult version of its AI model powering Movie Gen. Strike 3 asked for $359 million in damages and a permanent ban on Meta's use of its materials.
Meta denied all the accusations and has asked for the suit to be dismissed. The tech leader also claimed there is no evidence of it using adult images or video to train its AI systems.
"These claims fail not only for lack of supporting facts, but also because Plaintiffs' theory of liability makes no sense and cannot be reconciled with the facts they do plead. The entire complaint against Meta should be dismissed with prejudice," Meta stated.
Regarding the IP addresses being used as evidence, Meta said: “that the small number of downloads, roughly 22 per year on average across dozens of Meta IP addresses, plainly indicates that the videos were torrented for personal use.”
“With tens of thousands of employees and numerous contractors, visitors, and third parties accessing the internet at Meta every day, it's impossible to know who downloaded the clips or if they were Meta employees,” the company added.
“Strike 3 does not identify any of the individuals who supposedly used these Meta IP addresses, alleges that any were employed by Meta or had any role in AI training at Meta, or specifies whether (and which) content allegedly downloaded was used to train any particular Meta model,” Meta’s filing mentioned.
Meta slammed in its filing, quoting, “This presents yet another conundrum that Strike 3 fails to address. Why would Meta seek to ‘conceal’ certain alleged downloads of Plaintiffs’ and third-party content, but use easily traceable Meta corporate IP addresses for many hundreds of others?”
“The obvious answer is that it would not do so,” Meta argued, slamming Strike 3’s “entire AI training theory” as “nonsensical and unsupported.”
The social media company argued that the “torrenting began in 2018, before the company started researching multimodal models and generative video.”
Meanwhile, Meta started researching LLMs and generative video in 2022.
The tech giant was sued by a group of authors in 2024. Meta admitted to using the Book3 dataset, a 37GB compilation of 195,000 copyrighted books, to train its LLMs.
Meanwhile, Anthropic agreed to pay authors $1.5 billion over the alleged use of pirated books for LLM training in August. Apple was also sued by authors in September over claims it used a known body of pirated books to train its OpenELM LLM.
Also Read: Ethics of AI in Finance: Can Algorithms Be Trusted with Your Money?
“We don’t want this type of content, and we take deliberate steps to avoid training on this kind of material,” a Meta spokesperson told reporters.
For Meta, this controversy is not just about defending its reputation. It is a crucial test of trust. Winning the case would help the company reassert its commitment to responsible AI development and data transparency.