New Kindle Feature Uses AI to Answer Questions About Books
Key topics
The debate is heating up around Amazon's new Kindle feature that uses AI to answer questions about books, with commenters questioning the tech giant's rights to execute this feature and whether it's scanning and digitizing books to train large language models. Some argue that since users own the books, they're within their rights to feed the content into an AI, while others counter that users merely hold a revocable license to the books, complicating the issue. The discussion reveals a divide between those who see this as a legitimate use of user-owned content and those who suspect Amazon is exploiting loopholes to train its AI models. As commenters dissect the feature's technical details and potential implications, the conversation highlights the complex intersection of AI, copyright, and consumer rights.
Snapshot generated from the HN discussion
Discussion Activity
Very active discussionFirst comment
10m
Peak period
118
0-6h
Avg / period
16.6
Based on 133 loaded comments
Key moments
- 01Story posted
Dec 12, 2025 at 3:24 PM EST
26 days ago
Step 01 - 02First comment
Dec 12, 2025 at 3:34 PM EST
10m after posting
Step 02 - 03Peak activity
118 comments in 0-6h
Hottest window of the conversation
Step 03 - 04Latest activity
Dec 15, 2025 at 12:06 PM EST
23 days ago
Step 04
Generating AI Summary...
Analyzing up to 500 comments to identify key contributors and discussion patterns
Want the full context?
Jump to the original sources
Read the primary article or dive into the live Hacker News thread when you're ready.
what rights does a bookstore clerk need to answer questions about a product on the store's shelves? what a presumptuous question
I haven't seen a convincing argument why not. There's millions of librarians with the knowledge of more than 20 years of literature under their belt. Why can they answer your questions about a book but the robot can't?
Robots simply do not deserve the same consideration and the same rights that humans have
It's really that easy. Humans deserve more rights than inanimate objects
Companies like Amazon and Google have some really sticky fingers when it comes to intellectual property and personal data. I think it's worth asking these questions and holding them accountable for exploiting data that doesn't rightly belong to them.
That's what I mean by presumptuous. If that's really what they want the answer to, and what they object to, they should ask it plainly instead of alluding to it by asserting that there's some requisite but missing entitlement for the feature to exist on its face.
I mean they could have read it on company time as well.
However, let us not use a straw man here. Its not some company clerk, its one of the largest company on earth using other people's copy right to make more money for them selves.
That's my thoughts on that, anyway.
2. I doubt the Kindle version of the LLM will run locally. Is Amazon repurposing the author-provided files, or will the users' device upload the text of the book?
“Oh, you think you should be able to use an LLM with a book you paid for? Well you don’t own and book.”
Ok, and you like that? You want even less ownership? Less control?
I guess my argument is that Amazon shouldn't be able to have their cake and eat it too
I can totally understand that sentiment but I don’t think it’s logically aligned with wanting ownership of digital media.
We probably agree more than not. But users getting more rights isn’t universally good. To finish an argument, one must consider the externalities involved.
Hasn't training been already ruled to be fair use in the recent lawsuits against Meta, Antrhopic? Ruled that works must be legally acquired, yes, but training was fair use.
my favorite to eat is give other people my food, and have them tell me what not being hungry feels like
>My device, my content
I don't think you own the kindle store and servers used to train the Ai.
What do you mean? Presumably the implication is that it will essentially read the book (or search through it) in order to answer questions about it. An LLM can of course summarize text that's not in its training set.
Afaik, while the device is yours, everything else on it isn't.
Okay it's not 100% my device my content, so I shouldn't be allowed to run a local AI against the text?
But my opinion doesn't matter. Only Amazon's does. That's the point I was making. The premise of "my device, my content" is flawed and undermines the argument.
And I am not being cynical. That is literally what is on their web page, e.g.
https://www.amazon.com/dp/B0BTZT9PLM
https://www.theguardian.com/technology/2009/jul/17/amazon-ki...
In practice, that's not the case though, e.g. publishers on Kindle can choose not to allow text-to-speech assistive functionality.
In my experience, AI summaries often miss points or misrepresent work. There is a human element to reading a well written novel. An AI will miss some of the subtleties and references.
In any case, Amazon claims this feature is spoiler-free and that would be easy to implement. It likely works by feeding the book into an LLM context, and they could simply feed in the portion you've already read.
(This of course wouldn't be the case if they were reselling physical books.)
I suspect most of the people arguing this way would be in favor of more end user rights if we were talking about anything except the right to use AI.
“Rights good, AI bad” somehow leads to the insane argument that it’s a good thing you don’t have rights over the book you bought.
“You don’t really own the book” is a crazy argument unless the person saying this wants the locked-down DRM world where you can’t own a piece of media.
I am quite sure Amazon doesn't sell you that.
The amount of people completely - and likely intentionally - missing your point is both frustrating and completely unsurprising.
A quick reminder that this is part of HN's guidelines
> Please respond to the strongest plausible interpretation of what someone says, not a weaker one that's easier to criticize. Assume good faith.
"Yes this is a good question about 1984 by George Orwell, you could indeed be tempted to compare the events of this book with current authoritarianism and surveillance but I can assure you this book is a pure work of fiction and at best can only be compared to evil states such as China and Russia, rest assured that as a US citizen you are Free"
Welp. Seems perfect for a poison data effort !
As another example, I read the Aubrey-Maturin series earlier this year. Many times I would have liked a quick summary of a previous voyage or of a political plotline or something.
Don't be so judgemental.
I could see this being useful for that.
I can also imagine a character interaction graph, animated by chapter.
Oh, and pronunciation. The Sun Eater series is eloquent, but the names are inscrutable without having heard a few of the audio books.
Ok it's not just me that gets to the end of a page and it's like the page didn't exist.
On the other hand the times I use the search function on the ereader most are when I stumble across a continuity error. It would be interesting if a story-reading AI can be used to detect those. Not that I want there to be less human editing in books, if anything we seem to need more.
LLMs are great for this, for the plot and character questions, etc.
Authors have nothing to do with it. It’s my device, my book that I bought. It would be like if YouTube banned a screen reader. These are at two different levels of the stack.
The article links to a clear, direct counterexample of this claim. By Amazon, even.
https://gizmodo.com/fallout-ai-recap-prime-video-amazon-2000...
Amazon made a video with AI summarizing their own show, and got it broadly wrong. Why would we expect their book analysis to be dramatically better - especially as far fewer human eyes are presumably on the summaries of some random book that sold 500 copies than official marketing pushes for the Fallout show.
Why would that not also be true for the Fallout season one recap video?
I used to have to read fan wikis to figure this out.
But it will especially be useful for all the textbooks I’ve bought years ago. Being able to ask it questions (to the content itself) is better than asking ChatGPT or Gemini because they don’t have the content (they’re summarizing summaries found on the web)
I would much rather read a fan wiki than hope a LLM correctly understood a book's plot, at least with the current state-of-the-art of things.
Case in point: Amazon's own AI gets significant details of its own prestige TV show wrong:
https://gizmodo.com/fallout-ai-recap-prime-video-amazon-2000...
The Fallout fan wiki probably at least knows the Great War was in 2077.
You don't look up the end of a whodunnit before reading the beginning because that would make it kind of pointless.
> It also sounds as though authors and publishers were, for the most part, not notified of this feature’s existence.
This is perfectly reasonable fair use.
I'm starting to realize that a lot of content creators either don't understand fair use, or otherwise are unreasonable control freaks.
Well thats questionable actually
Sure the _result_ is transformative, but it had to consume the content in the first place to make a transformative. (grey area).
Yes, you _could_ argue that its a plain review, but you need to prove that its actually reviewing it rather than just quoting. But as its the machine doing it, that further muddies the water. Is it the end user whos generating the review? does the kindle license actually allow them to do that?
However, the other thing to note is that there is a contract between publishers and amazon that go over and above copyright. It will say how and where works can be distributed and how they can be processed. For example you're not able to distribute the book and then create your own audiobook version of it.
This is the lawsuit against the Diamond Rio all over again.
Afterall, I'd do the same thing, I don't trust amazon(the media part) with my IP
- https://github.com/omer-faruq/assistant.koplugin, which is forked from:
- https://github.com/drewbaumann/AskGPT
The first one even has prompts for quick recaps, summarize, translations, and more.
> I'm on page 750 of Anathem. Please give me a recap.
> You are currently reading the section of the book where the main characters have been launched into orbit aboard a repurposed military rocket and are preparing to board the alien starship, the Daban Urnud.
more recap details follow....
I would expect much more reliable results from chapter numbers though.
Just curious, not trying to attack you :)
Especially if I can get some sort of stats on the questions.
Like, if there were a lot of questions about when a character did something, then I know I wrote that badly.
Or if people talk about a set of characters, then I know that those characters made an impact.
Or if no-one asks about the book itself, but about some plot point or worldbuilding idea.
In general, if I, the author, can get a peek at this data too, it would be of immense value to me (and my publisher).
While I feel a certain amount of empathy to the authors, it's a table stake at this point to be honest.
Come to think of it, given how early O’Reilly had this it’s shocking to me that Amazon hasn’t done this sooner.
The O’Reilly Learning search was simultaneously the best and worst of all the early LLM applications. They have tons of high quality content that underpins very useful answers. I’ve also found a bunch of worthwhile books by looking through the sources.
It’s the worst because their template response is extremely unimaginative. I can be asking process questions about managing tech debt and it still gives me a code sample with every response as though I was asking “how do I add this button to my app”.