A federal judge has sided with the AI company Anthropic in its practice of training a chatbot on copyrighted books without permission from the authors.