It is a sign of the times that Amazon gets to call this fair use

Based on 15 sampled comments from this thread. Analyzed 2026-08-21.

Commenters spent less time on the headline’s legal framing than on the ugly incentive it exposes: physical-book ownership can let an AI company digitize and destroy copies while digital lending faces far tighter restrictions. Several readers see that as copyright law producing a perverse result. If a firm can lawfully buy and discard a book, scan it for model training, and retain the digital text, then preservation-minded libraries and archives are punished for trying to provide people access to the same material. Opinion split sharply over whether the destruction is actually meaningful. Defenders argued that most of these are low-demand, non-collectible volumes already headed for disposal, and that preserved scans are a better outcome than warehouse decay. They also treated learning from text as analogous to a person reading a book and forming useful ideas, making AI training a plausible fair use. Critics answered that digitization does not erase the loss of physical cultural artifacts or justify concentrating a corpus in Amazon’s hands. They distrust the claim that training is merely passive learning, pointing to models’ potential to emit protected passages and to commercialize knowledge without author permission. A deeper anxiety ran through the thread: proprietary AI systems could turn broadly available books into privately controlled training assets and mediated answers.

The core dispute is whether destructive scanning for AI training is an ordinary, defensible use of purchased books or a harmful extraction of cultural material. One side says low-value copies are routinely culled anyway, scans preserve their informational value, and machine learning from text resembles human reading. The other says commercial models can reproduce and monetize copyrighted work while the originals disappear, handing a concentrated AI vendor more control over knowledge rather than expanding public access.

• Praise: Digitizing overlooked, low-demand books for model training may preserve their informational content when physical copies would otherwise be discarded by sellers and space-constrained libraries.
• Praise: Several commenters regard AI learning from purchased text as the machine-scale counterpart of human reading and synthesis, not a forbidden act of copying ideas.
• Praise: The episode spotlights a fixable policy failure: extending first-sale-style rights to digital materials could let archives preserve and lend knowledge without forcing destructive physical workflows.
• Concern: Destroying print copies to feed proprietary AI pipelines feels like cultural loss disguised as efficiency, especially when a future reader may need the only surviving edition.
• Concern: AI training’s commercial payoff makes the fair-use analogy shaky for critics: a model can convert authors’ text into paid services and sometimes surface protected passages.
• Concern: The legal asymmetry lets a giant company retain digitized knowledge for private models while public-facing digital libraries are litigated into retreat.
• Concern: Readers fear that assembling vast book corpora behind corporate AI systems concentrates access to knowledge and makes future information increasingly filtered through platform-controlled answers.
• Concern: AI-generated material is seen by some as worsening cultural sameness, so sacrificing books to enlarge training datasets looks especially bleak rather than creatively regenerative.