Artificial intelligence is devouring data at an unprecedented rate, and now laboratories are forced to hunt for the last untouched sources of "human" content. This is not just about texts, but also faces, voices, and even movements—all of it is becoming raw material for a new generation of models.

Model Collapse: When AI Starts Eating Itself

The key problem facing the industry is the so-called "model collapse." This term, introduced into scientific discourse by a group of researchers led by Ilya Shumailov in 2023, describes the degradation of a neural network trained on data generated by other AIs. Each cycle makes the model more averaged, predictable, and detached from reality. It has been mathematically proven: mixing in synthetic data leads to the loss of rare events and the accumulation of errors.

The scale of the threat is confirmed by numbers. An Ahrefs study showed that more than 74% of new pages in Google search show signs of auto-generation. And according to Epoch AI estimates, the supply of usable training text on the internet could run out as early as 2026–2032. This is why books published before 2022 are becoming a goldmine: they are guaranteed to be free of "synthetic garbage" and "poisoned" data.

The Book Business: From Bulk Purchases to Lawsuits

The ISBNdb service, which aggregates metadata for 113 million publications, offered AI companies a specific service: purchasing books in batches of up to a million copies, followed by digitization and disposal. When the media wrote about this scheme, the company hastened to remove the page describing the service and stated that no deals had been made. According to its representatives, it was only a "test of market demand."

Significantly, the practice itself is not new. In August 2024, three American authors filed a class-action lawsuit against Anthropic, accusing the company of copyright infringement. According to case materials, Anthropic anonymously bought print runs of books, cut off bindings, and scanned them on an industrial scale. However, in June 2025, Judge William Alsup ruled that digitizing legally purchased books for training language models falls under the fair use doctrine.

Used-book dealers are already confirming anomalous demand: one seller's sales grew from twenty copies to several hundred per week. Buyers are not interested in genre or value—only the presence of an ISBN. Meanwhile, unique editions are lost forever, going under the scanner instead of onto a bookshelf.

Face Rental: The New Market for Digital Doubles

In parallel, a market for licensing appearances is forming. Platforms like ActID and New Claw pay people from $15 to $700 for the right to use their faces in AI series and advertising. In China, where more than 95% of the 128,000 AI mini-dramas released in the first quarter of 2026 were created using neural networks, the demand for real faces is enormous.

However, control over digital clones is proving fragile. British citizens who sold rights to their avatars to the startup Synthesia found their faces in propaganda videos supporting the dictators of Venezuela and Burkina Faso. These cases have become precedents for the actors' union Equity, which is pushing for a legislative ban on the uncontrolled transfer of rights to digital likenesses.

Not Just Texts and Faces

Robots also need examples of natural movements. Companies like Tesla pay up to $48 per hour to people who, in motion-capture suits, repeat routine actions to train humanoids. In 2025, more than $6 billion was invested in developing such machines.

The pattern is the same everywhere: previously, data was taken without permission; now, people are paid for it. But the structure of the deal does not change—the asset irreversibly leaves the owner's control, just now with a receipt.

My comment: The market for "clean" data is becoming as strategic a resource as oil or rare-earth metals. However, the industry ignores a fundamental risk: once rights to a digital copy are transferred, there is no turning back. Regulators should hurry to create clear legal frameworks, otherwise we will get a world where everyone can become a hostage to their own digital double.