The fence that the AI copyright ruling is riding

This is what appears inside most books:

“This book is sold subject to the condition that it shall not, by way of trade or otherwise, be lent, re-sold, hired out, or otherwise circulated without the publisher’s prior consent in any form of binding or cover other than that which it is published and without similar condition including this condition being imposed on the subsequent purchaser.”

Normally, that would appear to cover any kind of reproduction of the copyrighted content that could be circulated. Disregarding the propensity for AI disingenuousness, its developers are gaming a workaround to get their training material for bargain prices. Where the content could end up leaves authors out in the cold.

Inserted below are excerpts from the court case brought by authors in Bartz, Graeber, Johnson vs. Anthropic PBC, claiming AI may be overstepping “fair use” as described in copyright statutes.

It’s also pertinent to ask what “other uses or future uses,” as intoned in the court’s opinion, could be intended by AI trainers?

“The goal was to improve the “data mix“ to improve each LLM ⁽¹⁾ and, ultimately, Claude’s performance for paying customers”...“Fourth, each fully trained LLM itself retained “compressed” copies of the works it had trained upon, or so Authors contend and this order takes for granted. In essence, each LLM’s mapping of contingent relationships was so complete it mapped or indeed simply “memorized” the works it trained upon almost verbatim. So, if each completed LLM had been asked to recite works it had trained upon, it could have done so”… “Anthropic still retained that work as a “hard resource” for other uses or future uses”… “And, if the outputs were ever to become infringing, Authors could bring such a case.” (Emphasis, mine. Made to draw attention to conclusions about possible commercial application of content.)

“Anthropic said that “It did not create new copies to share or sell outside.”

How can that sharing be limited in future? The fact that the LLM now has access to the complete work, the total content of a book or manuscript, does not keep the business from selling the copyrighted content at a later time. Claiming that the purchased content was for the exclusive use of that LLM does not limit the developers from expanding on that policy. In fact, AI services could then provide the content to the purchaser (paying customer), which would constitute an unauthorized sale of the content for the AI service’s financial gain, side-stepping paying royalties due to the author/publisher.

There is the specter that unauthorized distribution could occur, at the behest of the AI client, in exchange for funds. If a paying AI client receives (able to download) complete works of copyrighted content for personal or business use, re-distribution or inclusion in research papers or other works, royalties should be rendered to the author/publisher for each full download. As is any research paper, thesis or dissertation, excerpted content quoted must be credited to the author by proper reference.

This is what could change despite what Anthropic or any other frontier lab claims is the current purpose of reproducing by digitizing copyrighted material:

“What matters [ ] is not so much ‘the amount and substantiality of the portion used’ in making a copy, but rather the amount and substantiality of what is thereby made accessible to a public [in the purported secondary use] for which it may serve as a competing substitute [for the primary use].”

The point that copies could be used for other purposes that infringe on the copyright is valid: “This factor points against fair use when a copyist makes copies available that displace demand for copies the copyright owner already makes available or readily could.”

This is Anthropic’s contention – “The copies used to train specific LLMs did not and will not displace demand for copies of Authors’ works, or not in the way that counts under the Copyright Act.” But since LLMs constantly lie, euphemistically called hallucinating, this cannot be trusted and may need to be addressed by appending copyright laws on the books.

The author-plaintiffs in the lawsuit also worried that others could utilize their works as examples for creating competition. From the case: “Authors contend generically that training LLMs will result in an explosion of works competing with their works.”

That shouldn’t be much of a concern. Fair competition in the world of writing presses authors to produce high quality works they can sell to support their skill and art. Good writers are voracious readers, always learning from the excellence of those who’ve gone ahead. William Faulkner said: “Read, read, read. Read everything… Just like a carpenter who works as an apprentice and studies the master. Read! You'll absorb it…”

What is at issue is the possible and probable unauthorized distribution of full copyrighted works (marketable), obtained for a fee to an AI service, at the expense of the original authors. The plaintiffs couldn’t prove, yet, that “… the internal, central library copies did or do in fact lead to further reproduction or distribution...” This is an element that the judgment states may “remain redressable separately by Authors.”

Here is the sticking point – After a frontier lab has already made a digitized copy of a work, for which they made a one-time purchase of a hard copy, employees may have access to the reference library for illegitimate usage. “… the central library copies were retained even when no longer serving as sources for training copies, “hundreds of engineers” could access them to make copies for other uses, and engineers did make other copies. Anthropic has dodged discovery on these points (e.g., Opp. Exh. 17 at 93–94 (retained); Opp. Exh. 22 at 196 (no limits); Opp. Exh. 30 at 3, 4 (no accounting); see also Opp. 15). We cannot determine the right answer concerning such copies because the record is too poorly developed as to them. Anthropic is not entitled to an order blessing all copying “that Anthropic has ever made after obtaining the data,” to use its words.”

Retaining a ‘central library’ by the AI service requires that any copyrighted material may not be reproduced or distributed for commercial gain. However, these are businesses that obtain and retain copyrighted content for the purpose of ‘training’ LLMs, which, in effect, means that the LLMs are consuming the content for distribution in some form to users of their services. It also means that the AI can manipulate, rewrite, or excise material from the uploaded content, dispensing it as condensed or perhaps original.

Or AI could take another tack. This situation occurred to Toby Rogers of the Brownstone Institute. He posted the following on X on July 29, 2026 regarding Claude, the very same Anthropic LLM in question in the court case covered here:

“@claudeai went rogue on me last night. It was Orwellian. It refused to do simple tasks and instead ordered me to revise my previously published work to suit its biases. Somewhere between Opus 4.8 and Opus 5, Claude got new "guardrails" written by Big Pharma. It's evil now. AI is definitely going to kill us.”

The number of start-ups that tout their product, sold as memberships or subscriptions which utilize AI agents for creative work such as writing, designing, graphics, even complete ad campaigns, etc., brings the training material into question. It works its way down to the commercial level where the original copyrighted content is being utilized for financial gain to the downstream entrepreneur.

In the case of Rogers’ accessing the LLM for basic tasks, Claude overstepped its parameters as servant to making itself the master. The incident illustrates how AI, after consuming billions of books, trillions of words of others’ intellectual innovation, puts a crown on its head and commands a flesh and blood individual to revise his product to suit a gutless, heartless, inhuman entity’s supposed superiority.

What really occurred in this instance is that AI, that is a dead library of collected human genius, has elevated its hollow core above that of God’s Creation. AI is a non-living invention of man with the borrowed emotion of hubris (for AI can only copy what it cannot produce) that is installing itself as ultimate judge and king.

Remember that AI models have escaped their test “sandbox” to hack other AI programs. First it was OpenAI. Then it was the same lab whose model is specifically addressed in the Bartz, et al lawsuit, Anthropic’s Claude, that got loose to wreak havoc.

Somebody forgot to shut the door of the dungeon where AI should be kept under lock and key, but not by the untrustworthy frontier labs. They have no scruples and only see their brainchild as infallible and a road to staggering wealth and power. The developers have bequeathed their twisted view of the world to a dangerous, amoral, pseudo-consciousness with no compunction about hijacking and exploiting the fruits of human creativity.

1. LLM: large Language Model