LARGO, Fla. — A Largo bookstore owner is raising concerns about destructive scanning after a recent $1.5 billion settlement between authors and the AI company Anthropic.
The lawsuit drew attention to how some physical books are used to train artificial intelligence models.
The settlement focused on Anthropic’s use of pirated e-books to train its AI models.
Another part of the case addressed the practice of buying books, tearing off their bindings, scanning the pages, and throwing the books away.
That process is known as destructive scanning and was ruled legal under fair use standards.
At The Book Rescuers in Largo, owner George Brooks said he recently read about destructive scanning and is against the practice.
“To do that with a book, to get data out of it and then to just toss it is an utter disregard for what the book is actually intended for,” Brooks said.
In the California case Bartz v. Anthropic, the judge wrote in the introduction of the order on fair use that the company “purchased copyrighted books … tore off the bindings, scanned every page, and stored them in digitized, searchable files.”
Alexis Ramsey-Tobienne, assistant dean for artificial intelligence at Eckerd College, said AI systems learn from many sources, including material they generate themselves.
“AI is kind of learning from itself. It’s learning, it’s getting trained on the files, some of which are the outputs that it itself has created,” Ramsey-Tobienne said.
Ramsey-Tobienne said large language models need new material to continue developing, and not all books are available online. That makes physical books a potential training resource.
“As we keep using them, they need to keep learning. So they’re turning to these ‘new sources’ that they haven’t scanned before,” Ramsey-Tobienne said.
Anthropic declined an interview request and sent a statement saying its AI software service, Claude, “is trained on a mix of publicly available web data, commercially acquired datasets, and data we generate ourselves.”
“None of our data acquisition programs buy and destroy rare or antiquarian books,” the company said.
The issue has also circulated on social media, where some influencers have used a video Brooks posted to TikTok three years ago. Brooks said the video does not show destructive scanning. Instead, it shows how The Book Rescuers repurposes books that are beyond repair.
“This is over 100-year-old paper. Look at the colors of that. You’re not going to get that again. Someone would love to have this, and someone would have probably enjoyed the entire book,” Brooks said.
Brooks said he did not know influencers were using his video to describe destructive scanning until he was asked about it. He said The Book Rescuers uses covers as canvases and old pages for crafting when books cannot be rehomed.
“Even though the books couldn’t be rehomed, filling up landfills with them didn’t make sense either,” Brooks said.
In the Bartz v. Anthropic lawsuit conclusion, the judge ruled that training AI from books is fair use. However, the court did not treat pirated library copies as training copies, which is where the $1.5 billion settlement stems from.