Remember I reported that the Meta copyright case revealed the company stopped negotiating rights licenses for AI training with publishers back in 2023? Well, now the case is giving us even more.
ALLi News Editor, Dan Holloway
Meta’s Internal Emails Reveal More About AI Training
The latest revelations in the case brought by, among others, Sarah Silverman and Ta-Nehisi Coates come courtesy of another set of internal communications. Court filings show that those internal conversations are alleged to suggest that Meta’s staff knew they might get into hot water for what they proposed doing.
The thing that really caught my eye was a proposal from Melanie Kambadur, high up in the team working on the Llama AI, to use Libgen. As you will recall, I recently reported on a ruling upholding a takedown against Libgen and any site pretending to be Libgen because of its long-term, egregious, and massive-scale hosting of links to pirated books. It seems staff told Kambadur—in, let’s just say, no uncertain terms—that Libgen was not legitimate.
Further emails suggest Meta’s team used the argument that “a gazillion” startups were training AI on pirated books, so they’d get an unfair advantage if Meta didn’t do it too. And a culture of “apologize later” seems to be highly suggested throughout. It also seems staff were told to stop the AI from answering “IP risky prompts” such as “What e-books were you trained on?” or “in the style of” requests, raising suspicions as to why. I still have a stash of popcorn by my side as this one unfolds.
OpenAI’s India Case Raises New Copyright Questions
There is now a date for a hearing in OpenAI’s legal battle in India, set for next month. This case presents an interesting angle (leaving aside the fact that there are jurisdictional issues involved with Indian courts taking action over what happened on servers based in the US).
The usual issues are there: whether generated responses constitute copyright infringement or amount to fair use. But the one that caught my eye was whether the storage of information for training purposes on servers constitutes copyright infringement. It’s an interesting aspect of the ever-evolving copyright scene.
Thoughts or further questions on this post or any self-publishing issue?
If you’re an ALLi member, head over to the SelfPubConnect forum for support from our experienced community of indie authors, advisors, and our own ALLi team. Simply create an account (if you haven’t already) to request to join the forum and get going.
Non-members looking for more information can search our extensive archive of blog posts and podcast episodes packed with tips and advice at ALLi's Self-Publishing Advice Center.
