{
  "id": 3762860,
  "title": "No need to destroy books in Singapore to legally use them for training AI, say experts",
  "url": "https://urgent.news/2026/08/27/no-need-to-destroy-books-in-singapore-to-legally-use-them-for",
  "topic": "ai",
  "section": "AI",
  "published": "2026-08-27T15:00:00.000Z",
  "source": {
    "name": "Straits Times",
    "slug": "straits-times",
    "url": "https://www.straitstimes.com/tech/no-book-shredding-needed-in-singapore-to-legally-use-them-for-training-ai-say-experts"
  },
  "original_language": "en",
  "account": "Singapore's legal framework permits the use of legally accessed works for artificial intelligence (AI) training, according to experts, eliminating the need for potentially controversial workarounds employed by companies such as Anthropic and Amazon in the United States. The assurance provided by this policy enhances Singapore's standing in AI and intellectual property law, as stated by Tony Yeo, vice-president of the International Association for the Protection of Intellectual Property's Singapore Group. Google's president, Kent Walker, emphasized the importance of such certainty for fostering innovation in the technology sector. A recent public consultation by the Ministry of Law and Intellectual Property Office of Singapore is exploring the impact of AI on copyright and patent regimes. While destructive scanning of books for AI training has raised concerns, a US court ruled that such practices, when transformative, constitute fair use of copyrighted material. Singapore's Copyright Act amendments from 2021 permit the use of legally obtained or accessed works for AI training, provided the source was purchased legally. Google's CEO advocates for clear rules allowing AI training on publicly available content, similar to regulations in Singapore, Japan, and the European Union. This approach would negate the necessity for developers to secure commercial licenses for every publicly available data point used in AI model training. Rights holders cannot exclude their published works from AI training, even if they indicate so. However, experts note that the Copyright Act does not permit AI models to generate responses that infringe on copyrights, such as producing an image resembling a company's copyrighted photograph of a sports star. Rights owners may deter AI training by restricting online access to their works, such as blocking downloads or keeping them behind paywalls. However, limiting access may reduce visibility and deter AI companies from licensing copyrighted works from individual creators or small businesses. Collective licensing by rights owners, as seen in the US, could ensure proper compensation for creators. Google's approach in the Google Books project, which digitized library books with care to prevent damage, demonstrates the importance of protecting rare books and respecting licensing agreements. The key takeaway from Google's perspective is the need for users to easily opt out of AI training if they desire, fostering a market with user choice.",
  "summary": "Tech firms in the US buy books, tear off pages to scan and feed AI models before shredding the pages.",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}