Meta allegedly used pirated books to train AI—US courts may decide if this is 'fair use'

Core Perspective - The article discusses the legal and ethical implications of AI companies, particularly Meta, using copyrighted materials for training their AI models, raising concerns among authors and publishers about intellectual property rights and fair compensation [2][3][5][24]. Group 1: Legal Challenges - Meta is facing a lawsuit in the United States for copyright infringement, with allegations that it used the LibGen dataset, which contains pirated materials, to train its AI models [4][10]. - The legal debate centers on whether mass data scraping for AI training qualifies as "fair use," a legal doctrine that allows limited use of copyrighted works under certain conditions [5][6][24]. - A significant case is The New York Times vs. OpenAI and Microsoft, where the newspaper claims its articles were used without permission for AI training [9][10]. Group 2: Industry Reactions - The Australian Society of Authors has called for regulations requiring AI companies to obtain permission and provide fair compensation to authors for using their works [13][14]. - Various licensing agreements are being established globally between academic publishers and AI companies to ensure creators are compensated while allowing data usage [21][22]. - The Authors Guild argues for a more favorable compensation model for authors, suggesting a 75% share of earnings should go to the author [15]. Group 3: Implications for Creators - The average median full-time income for authors in the U.S. was just over USD 20,000 in 2023, highlighting the financial vulnerability of creators in the face of AI advancements [12]. - The proliferation of AI-generated content poses a threat to original works, making it challenging to distinguish and protect intellectual property [16][17]. - As AI systems often do not cite sources, the value of attribution diminishes, further complicating the landscape for content creators [16]. Group 4: Regulatory Landscape - The European Union's Artificial Intelligence Act of 2024 aims to balance the interests of copyright holders with the need for innovation in AI, though its provisions are considered relatively weak [18]. - In contrast, the U.S. government has not enacted specific regulations for AI, with some officials arguing against excessive regulation [19][20]. - The Australian government has released a voluntary framework emphasizing transparency and fairness in AI systems, but no specific statutes have been enacted yet [23].