A.I. company Anthropic, the force behind popular language models such as Claude, was put under fire for destroying printed books by the mountain full in their new operation. The case, Bartz v Anthropic PBC, was decided at the end of July by the California district court. It concerned many issues, but connecting many of them was Anthropic’s “Project Panama”. The clandestine project involved the gathering of print books en-masse, chopping them up, scanning them into the A.I. engine, and then disposing of the remains (The Guardian, James). According to the ruling, the use of these books for training an A.I. model is legally fair-use, but some methods also used are more suspect, while the situation as a whole outwardly appears rather concerning.
Indeed, many of the actions taken by the A.I. company are causing a wide audience to react with dismay. Evidently, Anthropic was sure of that itself: in a memo, an Anthropic employee answered the question of why they would use the codename “Project Panama” to describe what they were doing, to which he answered by saying, “because we don’t want it to be known that we are working on this” (Court Listener).
It appears the methods Anthropic used were dubious from the start. The original goal was to improve their A.I. language models by training it on material written before 2022–before other A.I.’s started poisoning the sphere–and so they found no better solution than classic human works. The most straightforward option was e-books, which could be fed directly to the digital engines, but Anthropic reportedly couldn’t be bothered to try and deal with the texts’ original authors and forestall results. So, they quickly turned to pirated copies, a decision that landed them with a $1.5 billion court settlement they needed to pay. With online piracy becoming less appealing, a new option had to be considered: destructive scanning.
As allowed by the court, the company was able to take in whatever printed books they wished and feed it to an A.I. for its own learning, as such actions apparently did not violate copyright law. Such violations were only able to be avoided, however, because Anthropic was buying up books by the truckful, slicing open the spines, cutting down the pages, scanning the remains, and chucking it all into disposal once they were done (Ars Technica, Edwards). Since the books were legally purchased and then essentially tossed into the fire, the Judge ruled it to be free use. However, backlash built up as Anthropic’s slash-and-burn method of educating its A.I. models is destroying rare, and even unique physical copies of books. Many argue that the cultural ramifications of this are severe, and that without proper restraint this kind of destruction for the sake of A.I.’s progression can have serious effects upon the health of our global society. In the end, though some copyright claims managed to stick on the company, “Project Panama” has largely been given a greenlight to keep moving into the future, and it can only be hoped that other A.I. companies will not take this as a mandate to embark on their own shady operations, and destroy more precious physical media.
Author’s Note: I believe that the most telling part of this whole debacle is that the former Google employee that Anthropic hired to work on “Project Panama” admitted that they were trying to hush it all up. In my research, I have seen many tech-bros and A.I. apologists brush this case off without much concern. Many of them say that the books being destroyed are of no real value, or are at least an inconsequential lost. Also, from a property perspective, if you own a physical book you can do whatever you want with it, whether throwing it away, donating it, or destroying it. That is your right if it is your property, so long as you don’t make copies of it and distribute it without the copyright owner’s consent. However, I believe that condoning mindset is a slippery slope, and is giving too much slack to a force that will easily run away with it. A.I. companies, from their onset only a few years ago, have been incredibly acquisitive, competitive, and–most importantly–insidious. A.I. has crept into essentially every facet of life, and these companies are staking everything on staying ahead. So, with a mandate to destroy books and pour tons of physical information into the gaping maw of an A.I. blackhole, I worry it won’t be long until companies like Anthropic hold all the world’s information, and suddenly all of us are “Fahrenheit 451”-ing it, and only Claude will be able to tell us what “Fahrenheit 451” is even about. It is simply painfully ironic to see Anthropic–a company whose name literally means “of mankind”–destroying the totems of our culture and information on such a grand scale, and I strongly urge us all to stand against it.
Leave a Reply