What happens when tech giants rely on millions of books to train artificial intelligence but authors never gave permission The class action filed in 2025 by authors Grady Hendrix and Jennifer Roberson accuses Apple of using pirated copies of copyrighted books from the Books3 dataset to train its OpenELM models without consent or compensation.

Background: Why training data matters

Large generative AI systems depend on vast text corpora to learn patterns in language. Securing legitimate licenses for hundreds of thousands of books would be complex and costly. That gap has led some developers to rely on collections like the Books3 dataset which has been reported to include unauthorized digital copies of many in copyright works.

Allegations at the center of the lawsuit

• Dataset documentation Plaintiffs point to OpenELM materials that reference Books3 as a source, suggesting training data included unauthorized book texts.
• No licensing or payment Hendrix and Roberson say they never granted permission and received no author compensation for the alleged use of their works.
• Commercial harm The suit argues that using copyrighted books without consent undermines authors revenue streams and intellectual property rights.
• Class action scope The filing seeks to represent other authors whose works may have been used without authorization, raising broader industry questions.

Legal and industry implications

This case could set an important legal precedent about copyright and AI training data. A ruling for the plaintiffs may establish that using copyrighted books for training without permission can constitute copyright infringement. That outcome could require retroactive licensing deals and change the economics of AI development by making AI training data licensing and author compensation mandatory components of model building.

Tension between innovation and creator rights

The dispute highlights a key debate in AI policy The fair use defense for training data remains unsettled in court and legal experts are divided on whether large scale use of books for training qualifies as fair use. Greater training data transparency and clearer licensing practices are emerging as proposed solutions to balance technological progress with respect for creator rights.

Possible directions for the industry

• Adopt licensing frameworks so authors and publishers receive payment when their works are used to train models
• Increase transparency in model training data to allow audits and accountability
• Explore a books data commons that enables responsible access and compensates creators

Why this matters now

Generative AI is increasingly integrated into business and creative workflows. With roughly 42% of companies using AI to produce long form content, the outcome of the Apple lawsuit of 2025 could reshape how training data is sourced and paid for across the industry. For authors and publishers the case offers a chance to assert rights and seek fair compensation. For AI developers it could mean new compliance and licensing obligations but also clearer standards that support sustainable, ethical AI.

As the litigation progresses, stakeholders should watch for rulings that clarify the interplay between copyright law and AI training practices and consider building strategies that prioritize training data licensing, author compensation, and training data transparency.