On this episode of the Self-Publishing with ALLi podcast, Dan Holloway reports on Meta’s internal discussions about AI training, revealing that the company considered using pirated books to train its language models. Court documents show Meta abandoned licensing talks with publishers after realizing they didn’t have the rights to AI training data. Meanwhile, Kindle’s recent policy change regarding USB book transfers continues to stir debate, and World of Books expands into the US resale market.
Listen to the Podcast: Meta Considered Using Pirated Books for AI Training
Sponsors
All our ALLi podcasts are proudly sponsored by Bookvault. Sell high-quality, print-on-demand books directly to readers worldwide and earn maximum royalties selling directly. Automate fulfillment and create stunning special editions with BookvaultBespoke. Visit Bookvault.app today for an instant quote.
Self-Publishing News is also sponsored by book cover design company Miblart. They offer unlimited revisions, take no deposit to start work and you pay only when you love the final result. Get a book cover that will become your number-one marketing tool.
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.
About the Host
Dan Holloway is a novelist, poet, and spoken word artist. He is the MC of the performance arts show The New Libertines, He competed at the National Poetry Slam final at the Royal Albert Hall. His latest collection, The Transparency of Sutures, is available on Kindle.
Read the Transcripts
Dan Holloway: Hello and welcome to another self-publishing news podcast.
Clarification on the Kindle eBook Changes at Amazon
The timing of all things Kindle is on my mind because, obviously, last week's podcast was recorded before the story about the change in terms to Kindle went somewhat explosive, and obviously by the time this podcast comes out, the end date of February 26th will have happened.
I updated you all in my column in the meanwhile, and thank you very much to Howard for putting a little note on last week's podcast just to make sure that people were updated. I'll read that note again here. I will say, as I said there, ALLi has been in touch with Amazon. There has been some backwards and forwards, and I think, between us all, we have got to the bottom of it.
The situation is very much actually as it was originally stated in the article in The Verge, which is what I think triggered many of the slightly frantic posts I saw on YouTube, and what it is that is being removed from February the 26th, is the ability to download books to a computer and then transfer them onto your Kindle using USB.
You can still read the books that you've downloaded previously. You can still transfer across device that are Kindle enabled using Wi-Fi. So, it is just that USB transfer of downloaded books from your computer to Kindles that you can't do.
That said, although the change isn't as big as I've seen a lot of posts making out, it is nonetheless a big change, and it seems to be the case that the main driving force behind this, again, this is speculation, but it seems quite feasible speculation, is to do with a very old format of file that was usable on pre-2015 Kindle, and if you downloaded it and did the things that you can no longer do, then it was easier to strip out the DRM, basically.
So, this seems to be largely driven by DRM protection. So, there we go. There's everything brought up to date with the story around Kindle.
More Reveals from the Meta Court Case
That knocked out the order of the columns that I had anticipated last week, which means that we have two stories about the Meta court case appearing in quick succession, and I'll update you on the second of those now, because this does feel like it's the gift that keeps on giving as Meta is being sued around the training of its AI platform.
I think it's Llama that it's being sued in relation to, it's certainly Llama that the many documents that are now coming out in court relate to. Among those, bringing the case, Sarah Silverman. This is one of the very first cases I remember reporting on ages ago, Sarah Silverman's case, and Ta-Nehisi Coates. So, really high-profile authors bringing this lawsuit against them.
What we seem to have now is a lot of the internal comms within Meta are coming out in court, and on last week's episode, last week's cliffhanger as it were, what we had found out was that Meta had stopped speaking to publishers back in April of 2023, trying to get the rights to license authors works from publishers. The main reason they had stopped those negotiations was because they found out the publishers didn't have the rights, and the reason they didn't have the rights is they hadn't been having conversations with authors at that point about AI rights.
So, Meta decided it was too difficult to carry on those negotiations, and I had speculated wildly about whether authors knew that publishers had been hearing from Meta, and therefore might have been discussing rights with them at an earlier stage than I think we have subsequently found out in other cases around publishers and authors, was the case.
That's where we left off last week and this week what we have learned is what Meta did next. It seems that what Meta proposed they did next, that is, is basically not really object to a ‘do something and ask permission later' mentality.
So, emails from Melanie Kambadur, who is, it says in TechCrunch, high up in the team working on Llama. So, there are some really interesting things she proposed. One of the things she was proposing was using Lib Gen. Jen, as you will recall from a few weeks ago when the US courts confirmed the take down order that had been made against them because they are a giant repository of links to works, many of which are pirated. So, it is a known site, and it was known at the time that Melanie Kambadur proposed using them, that they were a renowned site full of pirated links, and still she proposed using them.
It's quite interesting. It is quite funny actually, reading the stuff that happened in court because it seems that the response that she got from one of her teammates was simply to send back a photo of the court ruling, the original court ruling against Lib Gen, which clearly demonstrated that they were pirates or hosted links to pirated books.
So, had clearly been thinking about using pirated books, and the argument they seemed to be using was, they used the word a gazillion startups were already training their eyes on pirated books, so if they didn't do the same, then those startups would have an unfair advantage.
There's no specific evidence that they did do the same, but that would've been their argument had they done the same.
That comes at the same time as internal emails seem to suggest that their AI was requested to be prevented from answering questions that were considered, and I quote, “IP risky prompts”, examples of which might have been, what eBooks were you trained on, or can you give me some text in the style of blah, where blah is a renowned author in whose style work could only be produced if the AI had been trained on their books, and their books as we know, they didn't have licensing rights to because of last week's episode.
Anyway, the ongoing saga continues, and I have to say I am finding it highly popcorn worthy.
World of Books Expands into US
Just a little story to end with. Well, it seems a little story because the Meta story, Meta and AI, and publishers and books, and who got the licensing rights when to what, is the big news at the moment.
World of Books. So, World of Books is a reseller. They sell second-hand titles, and they sell, what in the old days, we used to call remainder titles. I don't know if they still call them remainder titles, but I think the phrase I used in my column was titles coming towards the end of their useful period on the front list.
They buy these at a discounted price from publishers and then sell them on largely through Amazon marketplace. They have a huge turnover, $200 million. They're mainly based in the UK at the moment, and they are moving into the US. So, that is an interesting development in the e-tail trade of non-used books.
So, I will end with that, and we will see what delights we have waiting for us next week. I'm sure there'll be more on the Meta case. I'll keep my eyes peeled on that.
I am also looking forward to recording on Friday a session about European legislation around product labelling. So, I'm sure at some point you'll see a link coming up to that.
All good things, all exciting things to look forward to when I speak to you next.
With that, I will leave you and bid you a good spring, or for those of you in the southern hemisphere, a happy autumn. Thank you.
