AI companies are destroying rare books in order to train/grow their AI…

What was claimed

AI companies are destroying rare books in order to train/grow their AI models.

Our verdict

Needs caution

Reporting supports that some AI companies, especially Anthropic, have used destructive scanning and discarded physical books for training. But it is too broad to say 'AI companies' generally, because the evidence does not show this is a universal industry practice.

1 of 3 AI systems agree17 sources citedChecked Jul 30, 2026

Check your own claim

Paste any statement, headline, or AI answer — 3 independent AIs verify it in seconds, with sources.

Key findings

AI companies are destroying rare books in order to train/grow their AI models.

Misleading83%

All 2 AIs agree

Verified against 17 sources

Detailed Analysis

The core claim is broadly supported for Anthropic and some other AI firms: recent reporting says they have bought used and sometimes rare books, scanned them by destructive methods, and discarded the originals for model training. But the wording is too broad because it implies all AI companies are doing this, and the evidence only supports specific companies and reported practices, not the whole industry.

Why this verdict

The core claim is broadly supported for Anthropic and some other AI firms: recent reporting says they have bought used and sometimes rare books, scanned them by destructive methods, and discarded the originals for model training.
But the wording is too broad because it implies all AI companies are doing this, and the evidence only supports specific companies and reported practices, not the whole industry.

Claims checked

AI companies are destroying rare books in order to train/grow their AI models.

Misleading91%

This claim may mislead readers without added context.

What other AI think

AI companies have been buying large numbers of physical books, including rare, antique and out-of-print editions, before scanning their contents to train AI models and discarding the original copies. Anthropic purchased millions of printed books, removed their bindings, scanned every page and disposed of the originals as part of a project to expand the training data for its Claude chatbot. The claim is factually accurate based on recent reporting and court documents.

Why this verdict

AI companies have been buying large numbers of physical books, including rare, antique and out-of-print editions, before scanning their contents to train AI models and discarding the original copies.
Anthropic purchased millions of printed books, removed their bindings, scanned every page and disposed of the originals as part of a project to expand the training data for its Claude chatbot.
The claim is factually accurate based on recent reporting and court documents.

There is credible reporting that AI firms have purchased, scanned, and in some cases discarded large numbers of printed books to create training data. However the evidence that they are systematically destroying 'rare' books specifically, at scale, is weaker and often based on anecdote or inference rather than verified inventories. Thus the claim mixes documented practices with a broader, less-proven allegation about rare-book destruction.

Why this verdict

There is credible reporting that AI firms have purchased, scanned, and in some cases discarded large numbers of printed books to create training data.
However the evidence that they are systematically destroying 'rare' books specifically, at scale, is weaker and often based on anecdote or inference rather than verified inventories.
Thus the claim mixes documented practices with a broader, less-proven allegation about rare-book destruction.

Claims checked

AI companies are destroying rare books in order to train/grow their AI models.

Misleading75%

Multiple reputable reports show AI firms (e.g., documented court filings about Anthropic) bought, scanned, and discarded millions of printed books for model training, but direct proof that these actions targeted or destroyed rare/unique books at scale is limited and largely anecdotal.

This claim may mislead readers without added context.

Share this result

Try DoubleChecks