AI公司通过大量购买2022年前出版的书籍并采取破坏性扫描方式提取训练数据，事后将书籍销毁的做法引发舆论关注。[1]其中，Anthropic曾销毁数百万册实体书用于AI训练，该行动被美国联邦法院判决为合理使用。[1]
面对争议，马斯克随即宣布已要求SpaceX AI团队采用无损扫描方式。[1]与此同时，版权诉讼频频出现。7月14日，出版商对谷歌提起诉讼，称Google可能面临100亿至1000亿美元潜在罚款；[1]7月22日，新闻集团向加州奥克兰联邦法院提起反诉。[1]
AI companies have come under fire for purchasing millions of physical books published before 2022 and using destructive scanning methods to extract training data, subsequently destroying the books [1]. Anthropic has been identified as having disposed of millions of physical volumes through this process [1]. Despite the controversy, a U.S. federal court has ruled the practice constitutes fair use [1].
Following the public backlash, Elon Musk moved to distance himself and his company from the practice, announcing that he has directed SpaceX AI's team to adopt non-destructive scanning methods instead [1]. The scrutiny on AI companies' data acquisition practices has intensified, with multiple copyright lawsuits emerging against major tech firms. News Corporation filed a countersuit in federal court in Oakland, California on July 22 [1], while publishers initiated legal action against Google on July 14, claiming the company could face potential damages ranging from $10 billion to $100 billion [1].
