Why Would Meta Download So Much Porn?
A new lawsuit filed by Strike 3 Holdings reveals significant concerns regarding Meta's data acquisition practices for artificial intelligence training. Court documents allege that the company downloaded nearly 3,000 copyrighted adult films during its efforts to scrape material from across the internet.
The findings extend beyond copyrighted media. Evidence collected in the suit points to the acquisition of sensitive and restricted content including blueprints for 3-D-printable handguns, stolen password databases, and nonconsensual imagery. This discovery highlights the risks associated with the aggressive data harvesting models used by large tech firms to build their systems.
Legal filings suggest these files were acquired over a two-year window. The list of allegedly downloaded material includes images linked to the 2014 Celebgate breach alongside collections of deepfake content featuring various public figures. The inclusion of such illicit material in training datasets raises questions about the oversight and filtering processes Meta employs when gathering information from the web.
Meta has previously faced scrutiny for its reliance on scraped internet content and pirated books to improve its models. This latest legal action focuses on the specific types of data that end up in these massive repositories. As the litigation proceeds, the case will likely force a public examination of how the company vets the sources used to feed its artificial intelligence.