New Kindle feature uses AI to answer questions about books

Ownership, Licenses, and “My Device, My Content”

• One side argues that once a reader has paid for a book, how they process it (including with an AI tool) is their business; authors “got their money” and shouldn’t control reading methods.
• Others push back that with Kindle it’s only a revocable license, not true ownership, and Amazon’s DRM means “my device, my content” is factually wrong; Amazon ultimately decides what features exist.

Fair Use, Legal, and Contract Questions

• Some commenters call the feature “perfectly reasonable fair use,” likening it to a bookstore clerk answering questions or a reader writing notes/reviews.
• Others emphasize scale and automation: an LLM operating over the entire Kindle corpus is different from individual human reading, and training vs. inference is a legal gray area.
• There’s concern about whether uploading text to servers counts as distribution and whether publisher–Amazon contracts allow this kind of processing.
• A few point out recent rulings suggesting that training on legally acquired works can be fair use, though details remain contested.

Technical Implementation and Training Concerns

• Many assume the system won’t locally run; questions arise whether Amazon is reusing publisher files or user uploads.
• Several note that LLMs can answer questions by putting the book (or the portion read so far) into the context window at inference time, which is distinct from training.
• Skeptics doubt Amazon won’t also use this data for training, given its track record, and suggest “poisoning” Kindle-only junk content to pollute models.

Reading Experience and Target Audience

• Enthusiasts see it as extremely useful: recaps after long breaks, tracking minor characters, understanding dense classics, long fantasy series, textbooks, and generating study questions.
• Others deride it as a crutch for people who “hate reading” or “can’t be bothered to read properly,” arguing that forgetting earlier details is part of normal reading, or that this outsources the core experience.
• Some emphasize that people with limited time, long/complex books, or kids and jobs may genuinely benefit, comparing it to fan wikis and glossaries.

Accuracy, Hallucinations, and Alternatives

• Skeptics cite Amazon’s own faulty AI recap of its Fallout TV show as evidence that such systems can misrepresent works, especially with minimal human oversight.
• Supporters counter that text-based book Q&A is easier than video recap and should be more reliable if grounded in the full text.
• Several say they’d still trust well-maintained fan wikis over LLM interpretations for plot details and canon accuracy.

Authors’ Role and Control

• Some argue authors have no say in how readers navigate their books, even if it “spoils” mysteries or structure; others note that many works are carefully crafted for linear discovery.
• There is criticism that authors/publishers weren’t notified and can’t opt out. Others frame that as acceptable: this is a reader-side tool layered on top of legitimately licensed content.
• One perspective from an author is that aggregated question data could be invaluable feedback on confusing or impactful parts of a book.