Apple sued over using neuroscientists' books in AI training

Apple faces a lawsuit in California after two neuroscientists accused the company of using copyrighted books without permission to train its Apple Intelligence AI model.

Susana Martinez-Conde and Stephen Macknik, professors at SUNY Downstate Health Sciences University, filed the proposed class action on October 9.

They allege Apple trained its AI using pirated books from “shadow libraries,” including their works Champions of Illusion and Sleights of Mind.

The professors seek monetary damages and want the court to stop Apple from misusing their copyrighted work.

Apple is among several tech companies, including Meta, Microsoft, OpenAI, and Anthropic, that have faced lawsuits over unauthorized use of copyrighted content in AI training.

Last month, another group of authors also sued Apple on similar grounds.

Apple Intelligence is a set of AI features available on iPhones and iPads.

🔗 Source: Reuters

🧠 Food for thought

Implications, context, and why it matters.

Apple’s data gaps add compliance and liability risk

Apple’s technical reports go deep on architecture and benchmarks 1 2 3. They cite licensed material and public pages crawled by AppleBot with opt-outs, yet do not say how Apple filters pirated copies or avoids shadow libraries 2 3.
The US Copyright Office in May 2025 said training can be prima facie infringement when outputs echo inputs, and warned that using pirated or illegally accessed corpora counts against fair use 4. The complaint says Apple added more than $200 billion in value the day after unveiling Apple Intelligence, which could raise damages if training relied on unauthorized copies.
The case tests whether today’s disclosure norms work at trillion-token scale 3. Firms publish token counts and model design, yet keep dataset origins vague, so provenance gaps linger 3.

New demand for provenance-verified datasets and AI certifications

273 Ventures trained its KL3M model on 350 billion curated tokens from owned or licensed material or public domain text, and it reports better than expected results with law firms as clients 5. Common Corpus offers 500 billion public domain tokens 5.
Vendors can build marketplaces with pre-licensed corpora or vetted public domain collections 5. They can also offer provenance tools that flag shadow libraries and confirm lawful sources, with Fairly Trained signaling an emerging compliance market 5.
Investors track a shift away from blanket collective licenses the Copyright Office discussed but did not require 4. Capital is likely to flow to curation, licensing, and verification services that lower litigation risk.