Copyright conflicts between authors and AI remain complex

In early April 2025, about 100 authors from the Society of Authors (SoA) gathered outside Meta's headquarters (pictured) in London to protest the group's use of an online pirated book archive to train AI. The incident originated when the authors discovered that Meta was illegally using LibGen to train its AI model - Llama 3. LibGen is one of the world's largest "black libraries", storing more than 7.5 million books and 81 million unauthorized research articles. Authors Kazuo Ishiguro, Richard Osman, and Val McDermid discovered that their works appeared illegally on LibGen and filed a lawsuit.

Writer Val McDermid said: “In terms of law, there is a contractual obligation that when a third party uses a work, they must pay the author. Whether they adapt, translate or copy, they must accept payment because it is the author’s creative achievement.” Currently, members of SoA are not only protesting but also filing a lawsuit asking Meta to commit to respecting copyright, not continuing violations and compensating for damages. There are more than 14,000 signatures of support.

Meanwhile, a Meta representative said that they respected the intellectual property rights of third parties and that the way they used data to train AI was in accordance with current laws. The technology corporation said that they were following the law of “fair use” - a term in US intellectual property law that allows people to use works without the consent of copyright owners. However, this view is currently rejected by some courts in the US and other countries.

Earlier in March, Meta was also sued by a group of French authors. The publishing organization SNE, the author group SGDL and the musician group SNAC complained about Meta's behavior when using a series of copyrighted works without the permission of the authors and publishers. Christophe Hardy, head of SGDL, said he had filed a lawsuit with the Paris court, raising the issue that AI developers need to respect the legal framework. In the case of using copyrighted content, copyright fees must be taken into account. Meta admitted that it used a database containing the full text of about 200,000 books, including some in French, to train the AI ​​model - Llama.

In fact, copyright disputes related to AI are still ongoing and increasingly complex. Especially the issue of using unauthorized databases to train AI models. Authors' rights have been ignored, while the rapid and powerful development of AI is threatening content creators.

BAO LAM (Synthesized from The Guardian, Reuters)

Source: https://baocantho.com.vn/mau-thuan-ban-quyen-giua-cac-tac-gia-va-ai-van-phuc-tap-a185446.html