Google faces another AI training lawsuit from major publishers

Google and publishers are not strangers. Since the launch of Google Books, publishers authorized Google to scan books and provide snippet search services, establishing a framework of 'limited authorization.' However, this lawsuit alleges that Google used the same works—including copyrighted content obtained through Google Books and Google Play—to train the Gemini large model without permission. The core of the plaintiffs' allegation is 'scope of consent': authorization for search purposes cannot automatically extend to AI training.

The plaintiffs also cite a Google internal document that explicitly admits using copyrighted books for AI training is 'highly problematic for Google' and estimates potential fines between 10 and 100 billion dollars. This figure far exceeds the scale of the Anthropic fine, highlighting the risk Google faces. Note that the existence of the internal document is claimed by the plaintiffs, and Google has not responded to its authenticity.

Previously, California courts twice ruled in favor of AI companies, finding that using copyrighted works for AI training constitutes 'fair use.' However, both cases were in California courts, while this case is filed in the Southern District of New York. Different judges in different jurisdictions may reach opposite conclusions. Legal experts note that while California rulings are informative, they are not binding precedents, especially when case facts differ—for example, this case involves an existing contractual relationship between plaintiffs and Google, rather than mere web scraping.

Compounding the complexity is the Anthropic precedent: although California courts lean toward fair use for AI companies, Anthropic was still fined $1.5 billion for 'piracy'—the highest compensation in U.S. copyright history. In that case, about 500,000 authors each received at least $3,000, but many opted out of the class settlement to retain the right to sue over AI training. This shows that the boundary between copyright infringement and fair use is far from unified in judicial practice.

The plaintiffs' sharpest accusation is that Google 'intentionally removed or altered copyright management information (CMI)' to conceal the source of training data. If proven in court, this intentional act could constitute more serious infringement—the Digital Millennium Copyright Act (DMCA) has clear penalties for removing CMI. The plaintiff's complaint states that Google 'knowingly copied without authorization.'

Internal documents provide evidence of motive: if the copyright issue becomes public, fines could reach $100 billion. Although this figure is cited by the plaintiffs, if the documents are genuine, they would directly prove that Google executives were concerned about the legality of training. The timing of the lawsuit (July 2026) is also notable—filed shortly after the Anthropic fine verdict, publishers may have been inspired to take action.

This case could reshape the framework of cooperation agreements between AI companies and content creators. If the court finds that Google violated the limitations of its partnership, it will force AI companies to more strictly differentiate content usage scenarios and promote clear licensing terms. Current U.S. copyright law has not been updated since the Copyright Term Extension Act of 1998, leaving a legal blind spot regarding whether AI training constitutes derivative use.

Another possibility is that the lawsuit prompts legislative action by Congress. The tug-of-war between industry and copyright holders has lasted two years, but legislative progress outside judicial rulings has been slow. If this case proceeds to substantive trial, its verdict could become a bellwether for congressional action. However, given the lengthy litigation cycle, in the short term AI companies will continue to rely on the 'fair use' defense, while publishers apply pressure through class actions.