Eighteen advocacy organizations have asked the US Federal Trade Commission to investigate AI companies’ practice of buying physical books, scanning them for model training, and destroying the originals, according to The Register. The groups’ complaint frames the practice not only as a copyright-era flashpoint, but as a competition issue: once scarce or out-of-print books are consumed and discarded, rivals and the public may lose access to the same materials. The Register says fresh details emerged from document disclosures in Bartz v. Anthropic PBC, a copyright case brought by authors whose books were used to train Anthropic systems without permission. In those materials, Anthropic’s book-scanning effort was called “Project Panama,” and a 2024 internal memo described it as an effort to “destructively scan all the books in the world.” The same court exhibit, as described by The Register, said the project had a codename because the company did not want the work widely known. Employees were told not to discuss it in public areas or share it outside Anthropic. Anthropic did not respond to The Register’s request for comment. The mechanics are straightforward: physical books are acquired, digitized, and then destroyed rather than stored. The Register reports that older books are especially attractive for AI training because they are less likely to contain AI-generated text, which has begun to appear in newer written material. Destroying scanned copies also avoids the cost of warehousing physical inventory. The legal picture is unsettled. The Register notes that in the Bartz case, the district court accepted an argument that digitizing a physical book and destroying the original could support a fair-use theory by substituting an electronic copy for the physical one. The same article cautions that similar reasoning did not work out for the Internet Archive, underscoring that the doctrine is not a blanket permission slip for all scanning projects. The Register also reports that Anthropic is not the only company linked to the practice. Citing a recent report, it says Amazon has participated in book scanning and shredding; Amazon also did not respond to The Register’s request for comment. The provided cluster does not include that underlying report, so Amazon’s role remains a single-source claim here. Intermediaries are also facing scrutiny. The Register says ISBNdb reportedly helped broker book acquisitions for destructive scanning and recently removed a landing page advertising printed-book sourcing for AI large language model dataset needs. ISBNdb said the page had been part of exploring demand and that it had chosen to move away from that direction. The FTC angle matters because the groups are casting the issue primarily as anticompetitive conduct. Their argument, as summarized by The Register, is that large AI developers may be able to convert a finite supply of physical books into proprietary training assets while reducing access for other developers, researchers, libraries, and the public. That is a different claim from the familiar copyright dispute over whether model training requires permission or payment. For now, the story is a request for investigation, not an enforcement action. The next material fact is whether the FTC opens a probe, asks companies for information, or declines to take up the complaint. Who benefits: The Register says AI companies benefit by obtaining training text from older books while avoiding storage costs after scanning. If the FTC investigates, competing AI developers and public-interest groups could gain a forum for challenging those practices. Who's exposed: Anthropic is directly exposed because court disclosures described Project Panama, according to The Register. Amazon and intermediaries such as ISBNdb are also named in the report, though the specific Amazon claim is not independently corroborated within this cluster.