When Your Favorite Book Could Be the Last Copy Left

What happens when your favorite childhood book is the only copy left, and someone buys it just to destroy it? This could happen thanks to AI companies bringing you "the future."

My rare books

I own some old and rare books. I love books. The thought of shredding books, especially rare or old books, is just awful. I have a number of bookcases full of books that I was going to sell to pare down the collection. After what I learned about AI companies buying up rare books, and how they've destroyed books old and new, I'm rethinking it in case a rare book ends up in the wrong hands.

I have different books, like some old internet books from the 90's, that may not be "valuable," or even especially rare, but they preserve some history of the internet and are pretty funny to read now.

"The Internet Complete Reference" is just a funny title now, with a note on the bottom, "FREE Trial Offer! Buy this Book and Get 1 Month's Free Access on the Internet - Up to a $30 Value!"

Most of the older books on the shelf are technical books that still offer relevant knowledge even if implementations the exact package or tool being used to do something in a program have changed. I see a lot of companies claim to have "the next new thing," when it's actually a repackaging of a 20 or 30-year-old book on my shelf. I think that's worth remembering. Someone can claim they invented something if there's no more evidence of it.

I also have a few old medical books that are about 100 years old. I have a sociology book edition that's almost 80 years old.

I have old editions of classic books like The Odyssey, some novels from the 1960s for boys and for girls. To Dance, To Dream, and many Sociology books from the 1940's through 2010. And more.

Destructive scanning

A few days ago, I learned that Anthropic (the AI company who created Claude) destroys books. They cut them for fast scanning and then shred them. This post focuses on the Panama Project, which is Anthropic's project to digitally scan books (which they then destroyed to skirt copyright). Because of the favorable court ruling on that project, other companies seem to be following suit. OpenAI (the ChatGPT company) and Meta (the Facebook company) allegedly only used illegal downloads of digital books. OpenAI got in hot water for deleting the original data after models were trained, and before legal discovery.

These companies have all had to pay out for using these works without compensating authors.

Details came out of the Bartz v. Anthropic case about Anthropic taking books, cutting them for fast scanning, and then shredding them. The case was a copyright lawsuit in California where over 4,000 pages of court documents were unsealed, revealing internal plans like 'Project Panama,' Anthropic's initiative to scan 'all the books in the world,' destructively apparently. They also said in an internal planning document that was unsealed out of the trial: "We don't want it to be known that we are working on this."

Shredding rare books

Some of the reporting is saying many of the books Anthropic shredded came from used book sellers, but when I searched for more information, I saw articles about rare book dealers getting bulk purchases. They report that this is unusual. Someone looking for rare books will buy a specific copy or edition, or maybe a few of the same type of book. Not everything.

If companies are so callous with a regular book, how could we trust them with rare books? In fact, we can't. The Dallas Express reported in July 2026 that rare book dealers are watching editions disappear, some with only a handful of known copies remaining in existence. Destroying rare books is especially troubling, where a single book could be the last known copy.

Internet archive

Let me switch to a different organization that also ran into copyright issues. It was for different reasons, but the unfair incentives are real. The Internet Archive is a non-profit internet library. They're the source of the Wayback Machine, if you've ever used that. I mention them because they were also sued for scanning books (non-destructively), and being a library, sometimes a rogue one. While Anthropic's scanning was "transformative," the mission of preserving culture and sharing apparently wasn't good enough (Internet Archive lawsuit).

Anthropic's Claude will produce verbatim text from books that others use as their own words (Le Monde, AP, arxiv), while the Internet Archive lent books out digitally one at a time, some media only for an hour at a time, with DRM protection. Like a library. I don't know the details of the case, and I'm not a lawyer, but I do see an interesting contrast between the treatment of the Internet Archive and the treatment of Anthropic.

AI companies care about money

There's an incentive here for AI companies to destroy physical cultural heritage to avoid licensing fees, and treating books as disposable raw material. AI companies can apparently legally buy up, scan, shred, and share, while the Internet Archive, scanning for preservation and to be a library, was punished.

Books -> Computation -> Electricity

To process all the "raw material" of the books, including rare books, companies require massive computational power to create embeddings and train models. This is part of the drive behind AI data centers forcefully popping up around the country, including my neck of the woods.

Data centers are specifically cited by energy companies as the reason people's cost is going up or will go up. Gartner reported that data center electricity will grow 26% in 2026, with AI-optimized servers accounting for about a third of that. Power companies and counties say it results in strained grids and power bills go up.

Investors prop up companies until they're profitable

There's a reason this cost is being passed on (from big companies valued at billions of dollars, but paradoxically, aren't profitable). This is all being propped up by rich investors who can cover the costs of these companies until they might be profitable one day. Otherwise, we wouldn't be in this deep. AI companies are not making money (despite ongoing claims that "we will be soon, promise"). They're making billions but spending billions more to do it (OpenAI (ChatGPT), Anthropic ). Note the Anthropic article suggests they may have their first profitable quarter in Q2 2026, but don't expect to be profitable the rest of FY 2026. On top of that, they are valued at way more than any real number related to their businesses.

AI companies incentive is to push cost to you

Your electricity bills are going up to support all this because the money to support these efforts isn't there. That lack of profitability forces AI companies to push costs elsewhere. Some are starting to charge by token rather than by a monthly subscription. When people see the real cost of the AI they use, it's way more than you'd guess. Goldman Sachs reported that electricity prices jumped 6.9% in 2025 alone, which is more than double inflation. Goldman also reported that prices will continue to go up in that same article.

Here's my bill.

I don't know about you but I'm not sure I can afford that kind of increase. I'm in the mid-Atlantic region. In the past 5 years, my electricity bill went from around $80 - $100 a month, most months under $100. I think one heat wave it went up to $180. Now it ranges from around $200 - $450 a month. There was a heatwave in July so we're toward the upper range. I looked at our usage over the years and it's been the same. The fees and rates are increasing. My water bill increased similarly.

I am worried about me and my neighbors

This is not a NIMBY objection. My neighbors and my family are hurting from rising cost of living, which includes our insane utility bills. If these AI projects provided jobs or benefit to humanity, I could imagine living with it. But it's the opposite. This technology is unprofitable, harms jobs, harms companies, harms our ability to think for ourselves, harms learning, and harms creativity. It degrades the internet, and negatively impacts the environment. Furthermore, reports indicate that the associated data centers emit excessive noise and are causing adverse health effects for people who live near them. I don't want that for my neighbors in that town in my county, or the animals in the surrounding forest that will be left if they cut out a section for a data center.

Who controls information?

When those last rare books are destroyed for a billionaire's bottom line, that company controls that information. That's a piece of cultural heritage and history that can't be brought back. If we don't have history preserved in physical books that can't be censored, then we rely on these AI companies that are holding this knowledge.

They're telling us AI is the future. We don't have to go along with that. They don't get to decide our collective future, and I'm super proud of those high school grads who booed out-of-touch CEOs at their graduations earlier in the summer. Those CEOs are trying to push a future that people don't want. The future pushed back.

What can you do?

I typically like to try and include something practical you can do in my posts. I try to keep it simple, but I think this one deserves options.

The first one is you can avoid AI or use it intentionally. Other things are:

• You can use local AI models, even if your computer is not great. - SUPPORT YOUR LOCAL LIBRARY!
• Support writers, musicians, creators!
• Contact your representatives to express your views on AI, data centers, and more.

Don't use AI

Using fewer AI tools means slower workflow sometimes. It means paying subscription fees for human-run services instead of 'free' AI that can't summarize text without hallucinating. In my experience and reading AI slop articles online. The paid version I use at work can't do it well either. Similar to my opinion on how big tech is "convenient," NOT, AI is also "convenient" and "smart." People trust it to spit out news and correct information because it's convenient to do so. But it doesn't do either thing. It doesn't think or fact-check. It spits out language and fills in gaps with wrong information because we ask it for an answer when it isn't able to connect the dots for us.

I repeat. Don't use AI

What can you do? You can use your own brain for many things. I use my brain for ideas, writing, coding, system design, and more. It bums me out that AI might spit out my words for someone else to copy, but I would rather share what I know with others and risk it.

You can avoid giving AI companies training data from yourself. If you do use AI, make sure you read what it provides and question it. Assume you need to fact check it.

OK maybe a little AI

Some caveats to that. Because I don't mind being upfront with you.

A lot of people ask AI chats personal questions, medical questions, relationship questions, and more. If you can't afford a therapist, ChatGPT is free and the draw is hard to ignore. And if you can't get help from doctors, believe me, I know that experience. Depending on your condition, you can get desparate and upload a scan or lab results.

AI can cause irreversible harm and death

Be super cautious of what AI tells you. It tells some people things that happen to be helpful. It tells other people to hurt themselves and others. It can give very bad advice that have hurt people. Also be cautious because what you're saying and searching isn't private (ChatGPT, Anthropic and Google Drive). It could be leaked easily.

Don't help them train

When we provide conversations to AI companies, they're making money off us when we use their chatbots.

I realize that my blog is probably providing training data for AI. There are some things I could do to prevent this, but I did make the decision to allow it and hope AI recommends my blog and cites it. You might notice, if you looked at my html page sources (or the FAQ in each post), that I have some things optimized for AI chat bots to possibly recommend my blog. I do use local and encrypted AI chats to help find typos and check the flow of a post. I get a lot of activity that I'm pretty sure are bots crawling for info. But I do get visits that are from search engines and are probably real people. Like you. Thanks for reading!

You can use local AI, like GPT4All

There are local AI models you can use. Ollama is a popular way of using AI locally on a command line or terminal a text-based tool where you navigate your system and run commands by typing instead of clicking in a user interface . I've used the open source Software whose code is publicly available and modifiable. Transparency, collaboration, and often free as in free lunch are important parts of the open source model. platform GPT4All frequently. It can be downloaded here. I tested it on my old 4GB RAM laptop, and there are open source Software whose code is publicly available and modifiable. Transparency, collaboration, and often free as in free lunch are important parts of the open source model. AI models that fit on there and work.

Support your local library

I'll likely say "support your local library" as an answer to everything, but it feels especially important here. You can get all the entertainment, books, movies, music, and more from your local library for free, rather than pay subscriptions to companies that support AI hype.

Vote with your wallet

If you do use AI, I don't know if you can really avoid AI services built with copyrighted ingestion, but there are local models and privacy-respecting models.

Support writers and creators

Support creators directly. Don't upvote AI posts or low effort AI writing.

Contact your representatives and your local government

Let them know you don't support data centers.

Library image attribution: Michael D. Beckwith https://www.pexels.com/@2150568551/