Singapore can train AI models legally without destroying books, experts say

Legal specialists have weighed in on the debate surrounding AI training data sourcing, noting that Singapore's regulatory framework permits the use of published works for machine learning purposes without necessitating their destruction. The clarification comes as some American technology companies have adopted a practice of purchasing physical books, scanning their contents, and then shredding the materials—a method that raises both logistical and ethical questions. Experts suggest that Singapore's approach to copyright and fair use provisions offers greater flexibility for AI developers seeking training datasets. The distinction between Singapore's legal position and US corporate practices highlights varying interpretations of intellectual property rights and data acquisition methods across different jurisdictions. As artificial intelligence development continues to accelerate globally, the question of how companies source and utilise training materials remains contentious, with different regulatory environments offering contrasting pathways for compliance.