Legal Action Against Google Over AI Training Practices
Hachette Book Group, Cengage Learning, Elsevier, and author Scott Turow have initiated legal proceedings against Google in a federal court in New York. The lawsuit accuses the tech behemoth, based in Silicon Valley, of copyright infringement for improperly using copyrighted material to train its Gemini AI models.
In a nearly 60-page complaint filed on Friday, the plaintiffs contend, “Google willfully circumvented a long-established framework intended to safeguard copyrights, along with compensating authors and publishers, through a series of intentional decisions to develop Gemini1.”
The case asserts that Google illicitly obtained books as source material via Google Books, claiming the company misused works that were “secured for strictly limited purposes related to Google Books and other Google services.” Additionally, it is alleged that Google “downloaded scrapes from virtually the entire internet, including data from known pirate repositories and content sequestered behind legitimate paywalls.”
Furthermore, the lawsuit alleges that Google appropriated these works without authorization to train its AI models and continues to do so, operating outside existing contractual agreements.
The complaint emphasizes that the company was cognizant of the potential legal ramifications, noting that internal documents highlighted the use of books for AI training as “highly problematic,” potentially leading to fines amounting to $100 billion.
“At no juncture did Google notify authors and publishers that it was replicating their works as source material for the development and training of AI models,” the lawsuit asserts.
Kirk Sigmon, a founding partner specializing in technology and intellectual property law at KellDann Law, told Al Jazeera, “Any fair use argument that Gemini might advance could be undermined by the allegation that the books were unlawfully acquired.” He characterized the issue as complex, underscoring the difficulties in proving which materials were included in a training dataset.
This lawsuit follows an earlier attempt by Hachette Book Group and Cengage in February to join an ongoing class action initiated by a consortium of authors in 2023.
In a statement issued subsequent to the complaint, Hachette emphasized the collective determination of authors and publishers to defend their invaluable intellectual property rights across various genres, including fiction, nonfiction, children’s literature, memoirs, poetry, educational texts, and scholarly articles spanning myriad subjects.
Google has yet to respond to Al Jazeera’s request for comment.
A Surge of Legal Challenges
This case is not the inaugural lawsuit filed against an AI corporation by authors and publishers over alleged copyright violations.
A pending lawsuit against OpenAI includes notable authors, including George R.R. Martin, and the Authors Guild. In October, a federal judge dismissed OpenAI’s motion to have the case thrown out.
Conversely, an earlier lawsuit involving Meta, led by a group of authors including Richard Kadrey, did not favor the plaintiffs. In 2025, it was alleged that Facebook and Instagram’s parent company used copyrighted books to train its AI models, but a federal judge ruled that such training met the legal standard of “fair use.”
Michael Goodyear, an associate professor at New York Law School, commented on these copyright lawsuits, noting, “They typically follow a similar trajectory. The fundamental claims revolve around the unauthorized use of copyrighted works for training purposes. These are unlawful copies, and that forms the essence of the training argument. Some arguments also explicitly address infringing outputs.”
Nevertheless, Oli Huggins, CEO of ExpertEdge and VP of Partnerships at Packt Publishing, expressed concerns about the challenges faced in proving copyright infringement post-training.
He stated, “Establishing what transpires within a model presents another fundamental difficulty. Once the egg is baked into the cake, it becomes exceedingly difficult to pinpoint, quantify, or demonstrate which copy of a book was utilized. A model might exhibit familiarity with a work without reproducing sufficient text to establish the necessary evidentiary trail for a claimant.”
Huggins also criticized current licensing proposals as unsustainable for publishers, stating, “The economic model is deeply unattractive. Current offers can value a perpetual AI-training license at around $10 per title.”
This wave of litigation extends beyond literature, impacting various content-driven industries such as news and music, with numerous lawsuits against AI companies alleging copyright infringements.
CNN has filed a lawsuit against Perplexity, alleging that the company illegally copied more than 17,000 articles to train its models, resulting in content that was “identical or substantially similar” to CNN’s offerings, according to a complaint filed in May.
Recently, 17 news organizations, including The New York Times, accused OpenAI of withholding evidence in a case originally brought by the newspaper in 2023, asserting that the AI company infringed copyrights during the training of ChatGPT.
In the music industry, the law firm Hagens Berman has initiated a class action lawsuit against the AI music generator Suno, alleging that it trained its models on independent musicians’ works without consent.
In January, Universal Music Group also filed a lawsuit against Anthropic, alleging that the company infringed copyright by using 20,000 songs to train its Claude model without authorization.
Yet lingering questions persist about who bears ultimate responsibility for reproduced material alleged to have been copied, a conundrum that courts have yet to address. Goodyear pointed out, “If a user is actively attempting to induce the model to infringe, it could imply that the user is the one ultimately liable rather than the AI system. This remains an open question that the U.S. courts have not yet fully explored.”
Source link: Aljazeera.com.
