AI companies are shredding rare books

The observatory’s log for this cycle records a report circulating under the headline “AI companies are shredding rare books,” surfaced on Hacker News and shared via social media on July 27, 2026. The claim, as summarized, holds that firms engaged in training artificial intelligence systems have been destroying rare printed volumes in the course of acquiring material for their models.

Beyond the headline and its brief framing, the available record does not specify which companies, which books, or which archives are implicated. No primary account, dated incident, or named source has yet reached this desk to substantiate particulars. The report exists at present as an aggregated claim, discussed on a technology forum and relayed onward, without the documentary detail that would allow confirmation of scale or method.

What the material does establish is the shape of the concern being raised, that the acquisition of training data for artificial intelligence may carry a cost to physical cultural artifacts, and that this cost has drawn attention as a question of preservation and ethics rather than of algorithmic performance. Whether shredding here refers to literal destruction of hard-copy volumes after digitization, or to some other process, is not stated in the sources reviewed.

The Specola notes the report and its date of surfacing. Further observation awaits corroborating detail, additional sourcing, or direct statement from the companies named, none of which appear in the material currently on file.