- Zuckerberg would have prompted the implementation of the AI despite the employees’ objections
- Employees would have discussed the means to hide how the company acquired its IA training data
- Judicial deposits suggest that Meta has taken measures to mask its AI training activities without success
Meta faces a collective appeal alleging a copyright violation and unfair competition in the formation of its AI model, Llama.
According to court documents published by VX-Underground, Meta has downloaded nearly 82 TB of pirated book libraries Shadow such as the Anna, Z-Library archives and Libgen to train its AI systems.
Internal discussions reveal that some employees have raised ethical concerns in 2022, a researcher explicitly declaring: “I do not think that we should use hacked equipment”, while another said: “The use of hacking equipment should go beyond our ethical threshold. “
Despite these concerns, Meta seems not only to have plowed and take measures to avoid detection. In April 2023, an employee warned against the use of corporate IP addresses to access the hacked content, while another said that “the torrent of a corporate laptop does not feel Well, “adding a laughing emoji.
It is also reported that meta-employees would have discussed means of preventing Meta’s infrastructure from being directly linked to downloads, which raises whether the company knowingly bypassed copyright laws.
In January 2023, the Meta-PDG Mark Zuckerberg would have attended a meeting where he put pressure for the implementation of the AI to the company despite the internal objections.
Meta is not the only one to deal with legal challenges on AI training. Openai was prosecuted several times for having pretended to use books protected by copyright without authorization, including a case filed by the New York Times in December 2023.
NVIDIA is also under legal control for having formed its NEMO model in nearly 200,000 pounds, and a former employee had revealed that the company had scratched more than 426,000 hours of video per day for the development of AI.
And in case you have missed, Openai recently said that Deepseek had illegally obtained data from his models, highlighting the current ethical and legal dilemmas surrounding AI training practices.
Via Tom equipment
