Amazon’s Policy on AI-Assistance

Returned home tonight to find a post from Mark Dawson on Amazon’s Policy on AI-Assistance. The Amazon Terms of Service now contain statements on the use of AI to generate content in books.

The post is here in the KDP forum under Addition of AI Questions to KDP Publishing Process:

“…whether publishers make use of AI technologies or not, all books in our store must adhere to our KDP Content Guidelines, including the requirement that you accurately inform us when content is AI-generated.”

The post links to “help” in the KDP Content Guidelines:

“Artificial intelligence (AI) content (text, images, or translations)
We require you to inform us of AI-generated content (text, images, or translations) when you publish a new book or make edits to and republish an existing book through KDP. You are not required to disclose AI-assisted content. We distinguish between AI-generated and AI-assisted content as follows:

AI-generated: We define AI-generated content as text, images, or translations created by an AI-based tool. If you used an AI-based tool to create the actual content (whether text, images, or translations), it is considered “AI-generated,” even if you applied substantial edits afterwards.

AI-assisted: If you created the content yourself, and used AI-based tools to edit, refine, error-check, or otherwise improve that content (whether text or images), then it is considered “AI-assisted” and not “AI-generated.” Similarly, if you used an AI-based tool to brainstorm and generate ideas, but ultimately created the text or images yourself, this is also considered “AI-assisted” and not “AI-generated.” It is not necessary to inform us of the use of such tools or processes.”

More Questions than Answers

First question: how do I inform Amazon? Secondly what’s the burden of proof that content is or isn’t AI-generated? And in whose opinion?

Does a Generative-AI list of characters and settings make a work AI-Generated or only AI-Assisted? Says who?

If I outline an action scene and use Generative-AI to provide the full description of that scene including appropriate technical language, is that scene now Generated or Assisted? How do I identify whether that technical information and descriptive text is based on copyrighted works within the AI’s knowledge-base? Can I find that out?

Tarred with the Same Brush

What about book covers and other artwork? If I hire an artist do I have to supply Amazon with a list of tools they use to generate my covers? What about pre-made covers? I guarantee there are thousands on sale right now, created with AI over the last couple of years. Can I verify that? Can I verify that they aren’t based on copyrighted artwork within the AI’s model?

What about stock images? Shutterstock now has a filter for AI images. What if I don’t use it? And does Shutterstock guarantee full disclosure of image sources?

Photoshop now has an AI generative fill tool; if I use that to modify a stock image, is that image now AI-generated or AI-assisted?

There’s a significant rider in this section of the Help:

“You are responsible for verifying that all AI-generated and/or AI-assisted content adheres to all content guidelines. For example, to confirm an AI-based tool did not create content based on copyrighted works, you’re required to review and edit any AI tool outputs.”

Okay, I can review the output. It doesn’t give me the source or provenance of that output. It doesn’t cite references to source material in the AI model, or the copyright status.

Alice in Wonderland

This statement is nonsense:

“…to confirm an AI-based tool did not create content based on copyrighted works, you’re required to review and edit any AI tool outputs.”

How does my review and edit confirm the absence or copyrighted material in an opaque AI model? What edits can I make to confirm this? To what end? Do I change the punctuation? The word order? Cut a couple of sentences? Re-write it as Cockney rhyming slang? What’s the standard? Who adjudicates? How does this mitigate the fact that the AI generated the content in the first place? Do we re-visit the arguments about transformative works?

What happens if I can’t disclose because I don’t know?

While I can see this is Amazon’s attempt to clamp down on AI-generated content, these legal-sounding clauses make no sense and lack any objective measures.

So the next question is: whose account will Amazon shut down first? And who will take Amazon to court to test these new terms of service under the law?