Meta denies torrenting 2,400 adult films to train AI models

Meta has asked a US district court to dismiss a lawsuit alleging the tech giant illegally torrented pornography to train AI models. Strike 3 Holdings, an adult film company, filed the suit after discovering downloads of adult films on Meta’s corporate IP addresses, seeking damages potentially exceeding $350 million based on claims that Meta was secretly developing an adult version of its AI model powering Movie Gen.

Meta’s defense strategy: The company argues the lawsuit relies on “guesswork and innuendo” and lacks evidence connecting Meta to the alleged downloads.

• Meta claims Strike 3 Holdings “has been labeled by some as a ‘copyright troll’ that files extortive lawsuits.”
• The company argues there’s no evidence Meta directed downloads of approximately 2,400 adult movies or was even aware of the illegal activity.
• “These claims are bogus,” Meta’s spokesperson told Ars Technica.

Timeline doesn’t add up: Meta points to a critical gap between when downloads occurred and when its AI development began.

• The flagged downloads spanned seven years starting in 2018, about four years before Meta’s AI efforts “researching Multimodal Models and Generative Video” began.
• Meta’s terms of service prohibit generating adult content, “contradicting the premise that such materials might even be useful for Meta’s AI training.”

Personal use theory: Meta argues the evidence suggests individual employees downloaded content for private consumption rather than corporate AI training purposes.

• The activity amounted to only about 22 downloads per year—”a few dozen titles per year intermittently obtained one file at a time.”
• “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 filing stated.
• This contrasts sharply with lawsuits from book authors whose works are part of massive AI training datasets.

Network access complications: Meta highlighted the impossibility of definitively linking downloads to specific employees or purposes.

• “Tens of thousands of employees,” plus “innumerable contractors, visitors, and third parties access the Internet at Meta every day.”
• Strike 3 “does not identify any of the individuals who supposedly used these Meta IP addresses” or specify whether any were Meta employees involved in AI training.
• A Meta contractor allegedly downloaded content at his father’s house, but worked as an “automation engineer” with no apparent connection to AI training data sourcing.

The “stealth network” puzzle: Meta called Strike 3’s claims about hidden IP addresses logically inconsistent.

• Strike 3 alleged Meta used a “stealth network” of 2,500 “hidden IP addresses” to conceal some downloads while using easily traceable corporate IPs for others.
• “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?” Meta questioned.
• “The obvious answer is that it would not do so,” the company argued, calling Strike 3’s “entire AI training theory” as “nonsensical and unsupported.”

What they’re saying: Meta maintains its commitment to preventing explicit content generation in its AI tools.

• “We don’t want this type of content, and we take deliberate steps to avoid training on this kind of material,” Meta’s spokesperson said.
• The company suggested Strike 3 provided no evidence that Meta trained AI on its content because “there was none.”

What’s next: Strike 3 Holdings has two weeks to respond to Meta’s motion to dismiss, according to TorrentFreak, a news site covering file-sharing and copyright issues.