FTC urged to probe AI firms buying and destroying books for training data
Synopsis
A coalition of civil-society groups has asked the FTC to treat AI firms' bulk book-buying and destruction as an antitrust issue — not just a copyright one. The complaint, triggered in part by disclosures about Anthropic's Project Panama, argues that 'hoard-and-destroy' data practices could lock smaller AI rivals out of essential training material permanently.
Key Takeaways
More than a dozen civil-society organisations have formally asked the FTC to investigate AI companies' practice of buying, scanning, and destroying physical books.
The groups argue the practice raises rivals' costs and denies start-ups access to essential AI training material, potentially violating Section 5 of the FTC Act .
Anthropic's Project Panama — which involved destructively scanning millions of books for Claude training data — is cited as a key example; physical copies were recycled after scanning.
The complaint focuses on antitrust harm, distinct from ongoing copyright and author-compensation debates.
Rare, out-of-print, and specialised books are flagged as irreplaceable once destroyed.
The FTC has not yet announced whether it will open a formal investigation.
More than a dozen civil-society organisations have formally asked the US Federal Trade Commission (FTC) to investigate whether dominant artificial-intelligence companies are harming competition by purchasing, scanning, and physically destroying books used to train their models. The request, reported on 21 August, centres on what the coalition calls a 'destructive new data acquisition practice' that could entrench the market power of large AI incumbents at the expense of smaller rivals.
What the Coalition Is Alleging
The groups — including Demand Progress Education Fund, the Consumer Federation of America, and the Institute for Local Self-Reliance — want the FTC to determine whether acquiring and destroying physical books constitutes an unfair method of competition under Section 5 of the Federal Trade Commission Act. Their complaint is not aimed at restricting AI training itself, but at the competitive consequences of converting books into private datasets and eliminating the originals.
'Unlike a standard data acquisition strategy, this hoard-and-destroy practice could serve as yet another structural mechanism to raise rival companies' costs and deny start-ups and fledgling competitors a key source material essential to competing in the AI marketplace,' the letter stated.
The Anthropic Connection
The coalition's appeal follows disclosures that Anthropic acquired and destructively scanned millions of physical books as part of an internal effort known as Project Panama. Court records reported earlier this year indicated that book spines were removed so pages could be scanned rapidly, after which the physical material was sent for recycling. The company's stated objective was to obtain high-quality written material for training the AI models behind its Claude assistant.
Why Rare and Out-of-Print Books Are a Concern
The organisations raised particular concern about less common, out-of-print, and specialised books that may be difficult or impossible to replace once destroyed. They argued that large technology companies could deploy their financial resources to acquire books in bulk, effectively locking smaller AI developers out of the same source material.
'We urge the FTC to view this practice not in isolation but rather as the latest escalation in a documented pattern of anticompetitive conduct designed to create an insurmountable systemic moat around AI incumbents,' the groups wrote.
A New Competition-Policy Dimension
The coalition's complaint introduces an antitrust angle to a debate that has largely focused on copyright, licensing, and author compensation. An antitrust investigation would examine whether book acquisition and destruction exclude competitors — a distinct legal question from whether the training process infringes copyright. The FTC did not immediately announce whether it would open an investigation.
What Happens Next
The FTC has broad discretion over whether to act on the request. Should the commission open a formal inquiry, it could compel document production and testimony from major AI firms. The outcome could shape how AI companies source training data going forward, with implications for the broader competitive landscape of the industry.
Point of View
It sidesteps the murkier fair-use debate and targets something the FTC can act on more directly — exclusionary conduct. The Anthropic disclosures give the complaint concrete evidence rather than hypothetical harm. What mainstream coverage underplays is the irreversibility problem: once rare or out-of-print books are destroyed, no regulatory remedy can restore them. If the FTC declines to act, the precedent effectively permits incumbents to use capital advantage to permanently foreclose a shared informational commons — a structural harm that copyright litigation alone cannot address.
NationPress
22 Aug 2026
Frequently Asked Questions
What are civil-society groups asking the FTC to investigate?
They are asking the FTC to investigate whether dominant AI companies are harming competition by purchasing physical books in bulk, scanning them into private datasets, and then destroying the originals. The groups argue this practice raises costs for smaller rivals and denies them access to essential training material.
Which organisations are behind the FTC complaint?
The coalition includes Demand Progress Education Fund, the Consumer Federation of America, and the Institute for Local Self-Reliance, among more than a dozen civil-society groups. They jointly submitted the letter to the FTC on 21 August.
What is Anthropic's Project Panama?
Project Panama was an internal Anthropic initiative to acquire and destructively scan millions of physical books for use in training its Claude AI models. Court records reported earlier in 2025 showed that book spines were removed to speed up scanning, after which the physical copies were sent for recycling.
How is this different from existing copyright debates about AI training?
The coalition's complaint focuses on antitrust harm — specifically whether the practice excludes competitors — rather than on copyright infringement or author compensation. An FTC antitrust inquiry would ask whether the conduct unfairly forecloses rivals, a separate legal question from whether training on copyrighted material requires a licence.
Has the FTC agreed to investigate?
No. As of the date of the complaint, the FTC had not announced whether it would open a formal investigation. The commission has broad discretion to act or decline on such requests.
