Meta accused of using copyrighted materials for AI training

Artificial intelligence (AI) is becoming increasingly integrated into our daily lives, and as a result, more advanced models are being developed and trained on vast amounts of data. OpenAI’s GPT is probably one of the most well-known AI models, used in applications like ChatGPT. However, it’s not the only one out there. Meta has also joined the race with Llama, an open model that utilizes a diverse range of sources for training. But here’s the catch: the data used to train these AI models cannot be sourced without proper authorization, and it seems that Meta is now facing some legal troubles because of this.

A recent complaint was filed against Meta in a California court, accusing the company of using copyrighted data to train its AI models, specifically Llama. To make matters worse, it is alleged that Mark Zuckerberg himself approved this practice. According to the accusation, Meta allegedly used a tool called LibGen, known for collecting links from various sources, to gather data for training their models. However, among the data were numerous copyrighted materials, such as books available for download on the internet, which were also used to train Meta’s models.

Apparently, Meta had previously justified its use of LibGen for data collection, claiming that it had Zuckerberg’s permission. However, the new accusation goes even further, suggesting that Meta collected the training data without authorization and deliberately removed any references that could identify the source of the materials.

The current case accuses Meta of unlawfully using LibGen to collect data without taking the necessary measures to prevent the use of copyrighted content for model training.

This situation doesn’t reflect well on Meta, especially considering that even Yann LeCun, the head of Meta’s AI division, stated last year that authors should freely make their publications available for AI model training.

As of now, Meta has not provided any comments regarding the case.