Meta purportedly trained its AI on more than 80TB of pirated content ...
Meta is currently facing a class-action lawsuit accusing them of copyright infringement and unfair competition related to the training of their AI model, Llama. Court documents obtained from vx-underground reveal that Meta allegedly downloaded close to 82TB of pirated books from shadow libraries like Anna’s Archive, Z-Library, and LibGen to use for their AI systems.
Internal discussions within Meta showed that some employees expressed ethical concerns back in 2022, with one researcher outright stating, "I don't think we should use pirated material," while another employee mentioned that "using pirated material should be beyond our ethical threshold." Despite these reservations, Meta seemed to have continued with the practice and even took measures to avoid detection.
Concerns and Actions
In April 2023, an employee warned against accessing pirated content using corporate IP addresses, highlighting the ethical dilemma. Additionally, there were discussions among Meta employees on how to distance the company's infrastructure from direct links to the downloaded pirated content, raising questions about potential copyright law violations.
Industry Trends
Meta isn't the only tech giant facing legal challenges regarding AI training practices. OpenAI has also been sued multiple times for allegedly using copyrighted material without permission. Nvidia is another company under scrutiny for training its NeMo model on a substantial amount of books, and reports suggest that they scraped over 426,000 hours of video daily for AI development.
As the debate around ethical AI training continues, OpenAI recently accused DeepSeek of unlawfully obtaining data from its models, further emphasizing the ethical and legal challenges in the industry.
These revelations shed light on the complex landscape surrounding AI development and the ongoing efforts to ensure ethical practices and compliance with copyright laws.
Source: Tom's Hardware
