AI Continues To Steal Work, This Time It's Meta AI Stealing From 'Star Wars' & Authors

Tech

Mar 22

The latest in a chain of AI scandals has hit Meta, and its going to land them in a whole pile of legal trouble. They’re already in the midst of a lawsuit, (Kadrey v. Meta) where a group of US authors accused the social media group of copyright violations. However, the filings have revealed the scope of the potential breach. It appears that Meta’s language engine Llama was trained on information from Library Genesis aka LibGen.

For those not in the know, LibGen is a database of academic journals, graphic novels, nonfiction and fiction novels, all pirated, stolen from behind paywalls, and posted without the authors receiving a penny. The site boasts over 2.4 million non-fiction books, 80 million science journal articles, 2 million comics files, 2.2 million fiction books, and 0.4 million magazine issues. Star Wars authors, Michael A Stackpole, Timothy Zahn, Troy Denning, Kathy Tyres, and Brian Daley are amongst those pirated, as well as fantasy bestsellers, Stephen King, George R.R Martin, and Rebecca Yarros.

READ NEXT:

Featured

The Dark Side Of A.I.: How Artificial Intelligence Is Threatening The Future Of Writing And Art

The website is blocked (including its mirrors) in UK, France, Germany, Greece, Italy, Belgium, and Russia. However, the servers are known to be hosted in Russia and the Netherlands.

Authors now have a new search tool to see if their work appears on LibGen and, therefore, can add themselves to the legal action against Meta. And there will be an appetite in the writing community to do, because not only have none of the authors been compensated for their contributions, but there is a real risk that passages of their work could start appearing in the low-quality AI generated novels (and other content) that are currently flooding Amazon.

Whilst using other people's work for AI is questionable at best, this case is worse as Meta chose not to officially licence the works they were using, and thus generate revenue for the authors, but instead to gather the data from an illegal data store. Meta’s method of training their language module is not a victimless crime, and the scale of the lawsuit is likely to reflect this.

READ NEXT:

Featured