Amazon Launches New AI Policy: Self-Publishing News Podcast with Dan Holloway

As Amazon launches its new AI policy and the US Copyright Office refuses to register a copyright for an award-winning Midjourney artist, a consultation tries to untangle what is fast becoming the knotty mess of AI and copyright. Welcome to Self-Publishing News with ALLi News editor Dan Holloway, bringing you the latest in indie publishing news and commentary.

Find more author advice, tips and tools at our Self-publishing Author Advice Center, with a huge archive of nearly 2,000 blog posts, and a handy search box to find key info on the topic you need.

And, if you haven’t already, we invite you to join our organization and become a self-publishing ally. You can do that at allianceindependentauthors.org.

Listen to Self-Publishing News: AI Policy

Don't Miss an #AskALLi Broadcast

Subscribe to our Ask ALLi podcast on iTunes, Stitcher, Player.FM, Overcast, Pocket Casts, or Spotify.

OR, sign up to get notified via email right when a broadcast is about to go live on Facebook and when a new podcast is published (#AskALLi advice on Fridays and indie inspiration on Sundays).

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 Earlier this year 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: AI Policy

Dan Holloway: Hello and welcome to the new look Self-Publishing News podcast, if there can be such a thing as a new look for an audio format. Let's just call it the new format Self-Publishing News Podcast. So, for those of you who aren't aware, what we're doing now is we are breaking the news up so that there will be three separate news items across the week, and each news item will get its own place on the ALLi website which will enable me to bring you the news as it breaks.

I'm sure that some people, and yes I am yet again looking at you, Draft2Digital, will still manage to send me news so that it gets to you as late as it possibly can, but this will maximise the chances of getting the stories that matter to you most out in time, and also give me a little more space to expand on them on the blog.

Then the following week, I will be reflecting on those stories on the news podcast, and some of those you'll just get me talking to myself, talking to whatever version of myself I decide to talk to in order to make it slightly more interesting for you.

So, this past week has been, I guess it's a bit same old, but it's also been a really interesting week because we've had three really quite large stories, all of which have on AI, sorry for those of you who hate the AI, but they all have a really important AI angle.

Some of them will be of interest even to those who hate the thought of AI and the very mention of it, and the first one is going to be of interest to absolutely everyone because it's all about Amazon's new policy on artificial intelligence for works uploaded to KDP.

So, as we've seen quite a few announcements about AI policy in recent weeks. It started with journals, so online publications. So, Wired, for example had a really great policy for creators that it was never going to have AI-generated content taking up space on its column. AI-generated content would only be allowed where it was the story and was therefore sort of in quote marks.

Then more recently we had Kickstarter, who have gone a little bit more down the route of allowing AI, which is not surprising, but they still require there to be what they call the human input. So, provided you're open and transparent on Kickstarter about your use of AI, it's not a problem to have made some use of AI tools, but you are required to be in some way a creator.

Amazon's new policy is really interesting, because I'm not 100% sure that it goes even that far. So, what Amazon have done, and you'll be able to see this if you go to your KDP page, and you go to your guidelines on there, you'll see that there's a whole new section on AI content.

What that draws your attention to is Amazon's distinction between AI assistance, which is basically the use of tools like Sudowrite, anything where the AI is checking your grammar, checking your spelling, helping you in the creative process, polishing your prose, making the work shine, saving you time, that kind of thing; you don't have to declare your use of that, it's absolutely fine.

On the other hand, you have AI-generated content. So, that's stuff that comes, whether it's pictures through DALL-E or MidJourney, or whether it's text that has been generated by tools like ChatGPT. AI-generated content like that, you now have to declare. So, it's not banned, but you do have to declare it. So, that's It's interesting.

What's going to be really interesting is to see how Amazon enforce this. That is to say, is KDP itself going to have some kind of an AI-filter or some sort of filter that enables it to detect AI-generated content? If so, how on earth is it going to do that, because that would be really clever, and as I've reported on before, OpenAI have struggled with this. So, I'm not quite sure how they're going to do that. Are they going to rely on people's goodwill? I want to wish them the best if that's what they're going to do, because Amazon doesn't necessarily have a great record of people's goodwill being enough to stop the scammers.

So, how they are going to enforce that is going to be really interesting, and I'm sure we will have lots of news stories that focus on that in the not-too-distant future.

What's really interesting in this context is the way this fits in with one of the other stories that I reported on last week, which was around The Copyright Office and the Copyright Office's judgment on a piece of award-winning AI art created through MidJourney by the artist Matthew Allen.

This is something I reported on last year. Matthew Allen won an award at the Colorado State Fair for his artwork Théâtre de l'Opera Spatiale. So basically, it's the Theatre of Space Opera. It's really cool and looks like the cover of some sort of massively world-built Brandon Sanderson type fantasy sci-fi doorstop. As the title suggests, it's really good.

It won a prize, not surprisingly, and the artist, Matthew Allen, has sought to protect the copyright of this work, and just this month, the US Copyright Office has refused his latest appeal and said, no, you can't protect it. Which is really interesting, because while it is AI-generated, there are nonetheless some human touches. This would be stuff that gets through Kickstarter because, as was revealed, there are 624 human edits to the AI-generated art. Apparently that's still not enough to make it count as a human creation, because it's still fundamentally, according to the Copyright Office, generated by a computer, so it's not a human creation, it's not subject to copyright.

That's really interesting because it ties back to what we were just saying about Amazon, because if you're allowed to upload AI-generated content but AI-generated content doesn't enjoy copyright protection, then someone else can come along and treat your work essentially as though it were open source.

On the matter of public record, I see absolutely nothing that's stopping someone from just copying what you have written, putting their own maybe better cover on it, and uploading it as their own and selling that on KDP. It feels like it's something that hasn't necessarily been thought through consistently.

I can see people who are attempting to scam people by using AI, and themselves getting scammed by someone else who just comes along and copies it because it's not copyright protected.

All of which leads to another of last week's stories, which is that The Copyright Office, I think, recognizes that copyright law in the US is a bit of a mess when it comes to AI, and so it has launched a consultation, and you can submit to that consultation any time before October the 18th, and I highly recommend you do.

In the story I ran last week, there was a link to the consultation document. It's a really extensive 24-page document. It outlines what they are consulting on, and there are three really interesting areas that they're consulting on.

The first is one that we're all familiar with, how copyrighted works should be treated when it comes to training large language models, or any kind of large dataset machine learning program using them. That is, does the copyright owner have the right to determine how stuff is used, or can a tech company just buy a cheap copy of the book, or a picture of the art, shove it into the engine and then use it to train its own AI, and then commercial uses of that AI could be put into effect without any compensation for artists.

So, that's the first area. It's something that's been in discussion for a long time. The second area is the area we've already looked at, which is the copyright protection afforded to AI art, and what the Amazon story and the Matthew Allen story suggest is that it sort of makes sense to say that copyright applies to human creation, and there needs to be human creative input, but when it comes to practice, what does that actually mean?

As with so many areas of legislation and regulation around AI, I think this question, that sounds sensible, what does it mean, is going to be the key one in coming weeks, months and years, and you have a chance to put your opinion on that to this consultation.

The final area that's really interesting to me is what The Copyright Office should do about prompts that ask an AI to generate art that is in the style of someone, so mimicking an artist.

So, this isn't necessarily about misrepresentation; it's not about the kind of thing that Jane Friedman was scammed on, where someone just put her name on the cover of one of their books that contained total junk and said, this is a Jane Friedman book, buy it. It's not about that.

It's more about, it comes up quite often with art, with cover artists. There's a really famous cover artist whose style you really admire, and I want a cover in the style of blah, but I can't afford to pay a thousand dollars to blah, I'll get MidJourney to generate something in the style of them.

So, The Copyright Office wants to hear your opinions on how to deal with that. I mean, obviously this ties into the question of training data, because an AI is only going to know how to produce art in the style of someone else if it has been trained on what that someone else's style is. But it might be that again, there are questions of licensing fees, there are questions of your protection of, I guess the closest I could think of is trademark. Might it be that we invent a new, or The Copyright Office invents a new category that's not patent, it's not trademark, it's not copyright; it's some sort of protection for look and feel, for want of a better term.

Anyway, if you have feelings on that then do take part in that consultation.

The final piece of news last week that was related to AI was, I want to say a light-hearted piece. It won't be light-hearted for a lot of people. It was basically the Creative Commons art community has got together and issued an open letter basically saying, we love AI, we want to be part of deciding on its future.

So, this is in opposition to, or not necessarily in opposition to, but a counterpoint to the really big open letter that authors started after people had discovered that AI was, they thought, being trained on work they hadn't given permission to be trained on, and saying, this isn't on, we need better copyright protection.

The Creative Commons art community has basically said, no, AI is really good, it can do really amazing things. We want, as the Creative Commons community, to be part of shaping that future.

Obviously, the creative commons movement, many of you will have creative commons licenses for at least some of your work. It's a big movement that goes back well over a decade that has as its fundamental premise, the idea that open source is good, information should be open source, art should be open source, things that are created should be available for other people to use and base their own creations on, but you get to choose through setting up a Creative Commons license what those uses are.

So, as you will probably know, you can have a creative commons license that allows your work to be used by people, and you can state whether or not you want them to attribute you for your influence on their work, whether or not they can modify your work, and whether or not they can use your work as the basis for commercial content of their own.

So, the Creative Commons community want to be part of the drawing up of any legislation or regulation around that kind of argument around copyright. So, that's an interesting take.

This idea that these key parts of what we do with other people's work involve attribution, modification, commercialization, it actually gets to something pretty fundamental, and three of the key questions that we need to be talking about with copyright and AI.

So, there we are, that's last week's news which was really dominated by AI.

I imagine in next week's news, we will be talking about things less technological. We will probably be talking about the build-up to the Frankfurt Book Fair, for example, and no doubt lots of other things that haven't even broken yet.

So, thank you for listening to this first monologue, if you will, and I look forward very much to you joining me again next time. Thank you.