Quick Luddite Notes

🤖 Bits of Futurism

● AI Is Creating “Doom Loops” All Over the Economy — The AI industry is destroying its own incentive structure. (Oct 4, 2026)

Recently unredacted legal documents and depositions in the New York Times’ lawsuit against Microsoft and OpenAI revealed something extraordinary: executives knew full well of the destructive powers of AI when they rushed to launch products featuring the tech in the wake of the explosive success of OpenAI’s ChatGPT.

Internal Microsoft documents quoted by the newspaper’s lawyers warned of an ugly reality now playing out across the entire economy: “doom loops,” in which AI monetizes the work of human creatives while undermining the economic systems that funded their work in the first place.

“Our AI content strategy has started a ‘doom loop’ that will hurt the performance of our models and the entire web at the same time,” one internal Microsoft document reads. “It is highly unusual that an end-product threatens the economic foundations of its essential suppliers, but that is the situation we have created for our LLM business with respect to its ‘content supply chain.'”

The closer you look, the more you see doom loops everywhere. Journalists, for example, used to publish work under the assumption that online platforms would drive readership to them. But now, tools like ChatGPT and Google’s Gemini simply summarize their articles without sending readers to their sites, sparking a crisis that seems poised to wipe out the independent media.

Similar stories are playing out across other industries, from music to filmmaking. Writ large, many are concerned that AI will eventually be strong enough to wipe out huge numbers of jobs, creating a situation in which unemployed consumers won’t have the funds to support the economy.

The irony, as the Microsoft document pointed out, is that the AI industry is a snake eating its own tail: the very products that it’s pushing on users rely on fresh cultural content even as they destroy the economic incentives to create it.

...

Worse yet, as the legal documents in the NYT‘s copyright lawsuit showed, insiders admitted that the entire operation is built on theft.

“Millions of people around the world will soon consider large models ‘hoovering up’ all their work to be an astonishing theft of unprecedented proportions,” an internal Microsoft document admitted.

● Court Finds That Using Copyrighted Material as AI Training Data Is Not Fair Use — Could this be the moment copyright holders have been waiting for? (Oct 4, 2026)

The “fair use” doctrine under the US Copyright Act of 1976 allows for copyrighted materials to be used for purposes like criticism, journalism, and research.

Whether something is deemed fair use depends on several factors, including how transformative it is, how much of the work is used, and whether the reproduction alters the market value of the original.

It’s the keystone of the AI industry’s legal defense in a litany of ongoing lawsuits filed by rightsholders, who are accusing tech companies of unlawfully using their intellectual property to train their AI models.

...

Now, the legal system is starting to show signs that the defense could be on thin ice. Last week, a US appeals court upheld a ruling that AI legal research company Ross Intelligence had broken copyright law by training its legal search engine on materials taken from Thomson Reuters’ Westlaw legal research platform and database.

Ross shut down in 2021, shortly after the lawsuit was filed, citing mounting financial strains caused by the litigation.

The decision by the Third Circuit court of appeals — a first-of-its-kid ruling, per Reuters — marks a pivotal, albeit heavily nuanced, moment in the ongoing fight between AI companies and rights holders, who accuse the former of unfairly taking advantage of their work.

There have been dozens of other lawsuits like it, filed by authors, comedians, the music industry, influential newspapers, and even Encyclopedia Britannica. For years now, the pressure has been mounting as more and more rights holders cry foul, accusing AI companies of plundering their work.

How influential the latest appeals court ruling will be is debatable. The court found that Ross had effectively copy-pasted Thomson Reuters’ “headnotes,” or brief editorial summaries of legal issues, verbatim for its legal search engine, a practice that isn’t difficult to separate from fair use.

“Ross took the headnotes to make it easier to develop a competing legal research tool,” a Delaware federal court judge wrote in the original ruling last year. “So ​Ross’s use is not transformative.”

In fact, as Copyright Lately’s Aaron Moss points out, even the AI industry called the latest ruling a win, with tech industry group Chamber of Progress senior director of AI Adam Eisgrau tweeting that the ruling “implicitly confirms that highly transformative gen AI training to produce a hugely multi-purpose model with substantial public benefit is likely fair use!”

“Judging by all the victory laps, you might think everyone had read a different opinion,” Moss wrote. “They hadn’t.”

I beg to differ on several grounds. Nothing is relevant in the above report.

The case opened in 2021 couldn’t have been about AI, or generative AI. LLMs were not a thing back then. Either way, copy-pasting stuff, just like a search engine’s use of excerpts of news reports (a hugely disputed topic in countries like Australia and Canada), has nothing in common with “using copyrighted material as AI training data.”

Then, I very strongly object to the idea that using copyrighted material to train an LLM would constitute intellectual property infringement. An LLM would absolutely never reproduce verbatim any of the texts used for its training! Never ever! An LLM “learns” by creating weights and biases. It creates correlations much like a living creature learns. However, unless it’s using a Retrieval-Augmented Generation (RAG) or similar system, an LLM won’t cite anything exactly “from memory.” So, for all intents and purposes, this is transformational!