Apple is once again in legal trouble over its artificial intelligence training practices, with two US neuroscientists filing a proposed class action lawsuit accusing the tech giant of using pirated material to train its AI models.

The complaint by two professors at SUNY Downstate Health Sciences University alleges that Apple trained its Apple Intelligence AI using “shadow libraries” (online repositories of illegally copied books) including their own copyrighted works.

The neuroscientists claim Apple accessed these databases through web-scraping tools to train models such as OpenELM, without seeking permission or offering compensation.

The lawsuit references the Books3 dataset, a massive online collection of more than 190,000 pirated books that has been at the centre of similar lawsuits against Meta and Anthropic.

This marks the second lawsuit in just over a month accusing Apple of copyright infringement in AI training.

In September, two authors filed a similar claim, alleging that Applebot, the company’s web crawler, pulled data from the same shadow libraries.

Apple joins a growing list of tech giants facing scrutiny for how they source material to train large language models.

OpenAI, Meta, and Anthropic have all faced lawsuits over similar claims, with Anthropic recently agreeing to a US$1.5 billion settlement to resolve allegations from half a million authors.

The dispute highlights a growing tension between copyright law and the data-hungry nature of AI development. While companies argue that training on copyrighted material falls under “fair use”, courts have yet to deliver a definitive ruling.

Apple has not yet commented on the lawsuit.