Chinese tech giants linked to pirate e-book scraping for AI

A new report alleges major Chinese technology companies utilized pirated e-book repositories to train large language models, raising significant intellectual property concerns. The investigation claims these firms systematically scraped copyrighted text from unauthorized shadow libraries to build training datasets. While specific companies deny wrongdoing, the allegations highlight persistent challenges in verifying AI training data provenance. Copyright holders argue this practice constitutes industrial-scale infringement that undermines licensing markets. Legal experts note that using illicit data could expose developers to liability under both Chinese and international copyright frameworks. The findings arrive as regulators globally scrutinize AI supply chains for compliance gaps. Industry stakeholders are now debating whether technical audits can effectively detect pirated content in massive datasets. This controversy may accelerate legislative efforts to mandate training data transparency and record-keeping requirements for AI developers operating across jurisdictions.