一份关于5000本冷门书目的“荒唐”订单引发了怀疑
A 'bananas' order for 5000 obscure book titles fuels suspicion

原始链接: https://www.irishtimes.com/world/europe/2026/08/10/a-mysterious-buying-spree-is-unsettling-europes-booksellers/

从高威到柏林,欧洲各地的独立书店纷纷报告称,收到了一大批可疑的、大规模的在线订单,内容多为内容驳杂且通常冷门的书籍。零售商怀疑这些采购是由人工智能驱动的机器人策划的,旨在获取纸质书以“喂养”大型语言模型。 这种做法与人工智能公司Anthropic的“巴拿马计划”(Project Panama)如出一辙,该计划涉及购买、扫描并销毁数百万本书籍,以训练其Claude聊天机器人。尽管Anthropic否认使用了稀有或古籍,但欧洲书商认为,第三方供应商正在整个欧洲大陆搜集书籍,以规避欧洲更严格的版权法,并很可能将其运往“合理使用”政策更为宽松的美国。 行业专家认为,这种大规模扫描并将其制成纸浆的操作既具有破坏性,也侵犯了知识产权。除了法律顾虑外,书商还担心这些做法的长期影响,从对独立零售商的生存威胁,到对人类批判性思维的潜在侵蚀。许多书商现在拒绝参与其中,呼应了高威书商托马斯·肯尼(Tomás Kenny)的观点:如果目标是未经授权地攫取知识产权,他们绝不参与。

《爱尔兰时报》近期的一篇报道指出,一笔涉及5000本冷门书籍的巨额订单在Hacker News上引发了关于人工智能训练与实体媒介之间关系的争论。讨论的核心在于:为进行数字化扫描而大规模购入书籍——即通过拆除书脊来实现数字化——究竟是弊大于利,还是属于数字保存的一种形式。 评论者对此意见不一:有人认为,既然这些书籍中的许多本已面临化浆处理的命运,那么将其用于人工智能训练反而是一种改善,前提是由此产生的数据能够公之于众。另一些人则对可能存在的珍稀或绝版书籍被毁感到担忧,认为这是一种将企业数据收集置于文化保存之上的“版权洗白”行为。 部分参与者批评了对这类订单的怀疑态度,认为书商应当欢迎这些收入;而另一些人则将这种激进的“训练材料最大化”与“回形针制造机”的思想实验相提并论。归根结底,这一讨论凸显了关于人工智能发展伦理、实体档案脆弱性,以及使用受版权保护材料进行机器学习所带来的法律模糊性等深层次的矛盾。
相关文章

原文

From Galway to Berlin and beyond, independent booksellers across Europe agree that something very strange is going on.

Last May Tomás Kenny of famous Galway book retailer Kennys received what he describes as a “bananas” online order for 5,000 books. But it was less the scale of the order that caught his eye and more its eclectic selection.

“Some was high-quality non-fiction, like A History of Connemara, and the next thing might be The Eddie Hobbs Guide to your SSIA,” said Kenny.

Large orders are nothing new for the kennys.ie website. It was only the second bookshop in the world to launch a website in 1994 and it is popular with large institutions, such as Asian universities seeking to stock shelves in newly opened English libraries.

But when publicly funded institutions contact him, they usually negotiate a bulk price. May’s online order – with no haggling – sent up a second red flag.

Last May, Tomás Kenny of Kennys Bookshop, received a 'bananas' online order for 5,000 books

Over in Berlin, booksellers in the German capital are grappling with similar-sized mass orders. They arrive in the night via email, are as profitable as they are perplexing, involving dusty books that no one in their right mind would buy. Pass Your Driving Test, 2018 Edition, anyone?

Booksellers say it is not unusual for these orders to range from €10,000-€50,000, welcome income amid an ongoing earthquake in their industry thanks to Amazon and other online booksellers. Yet many fear that, with these bulk orders, they may be signing their own death warrant.

As evidence, they point to a January Washington Post report about a similar mass-order phenomenon among US antiquarian book retailers.

According to that report, AI giant Anthropic was behind Project Panama, spending “tens of millions of dollars to acquire and slice the spines off millions of books, before scanning their pages to feed more knowledge into the AI models behind products such as its popular chatbot, Claude”.

Anthropic’s Claude AI models hack into 3 outside groups during testingOpens in new window ]

Details of the US plan emerged through filings related to a copyright lawsuit against Anthropic. That ended with a $1.5 billion settlement with book authors.

Berlin antiquarian book dealer Urban Zerfass has no doubt some tech giants are engaged in a similar practice in Europe.

“The Anthropic plan is clear in documents to its investors, meanwhile Amazon earns 20 per cent on every transaction,” he said.

Germany’s Publishers and Booksellers’ Association has described mass scan-and-pulp practices as destructive and illegal.

“Under German law scanning books – regardless of the purpose – would not be permissible and constitute a clear violation of copyright law,” said Thomas Koch, spokesman for the association. “That this is now happening with second-hand books is another highly troubling example of this practice.”

German book retailers have put their heads together and suspect the entire ordering process is being run automatically, by an AI-driven bot, with a focus only on books with an ISBN number, the identification system introduced in 1966.

AI giant Anthropic was behind Project Panama, spending 'tens of millions of dollars' to acquire books and scan their pages to feed more knowledge into AI models behind products such as its popular chatbot, Claude. Photograph: Jason Henry/The New York Times

“Today the ISBN allows the bots or AI log what books it still lacks or has yet to scan,” Markus Brandis, art historian buyer at Berlin auction house Bassenge, told Germany’s Welt daily.

What is happening and its legality depends, however, on where it is happening. European booksellers say they have noticed a shift in the mass orders, away from costly shipping of titles across the Atlantic to addresses around Europe. The suspicion is that these are collection points for books, which are then shipped in bulk in containers to the US.

There, more liberal “fair use” policies make it legal for the owner of a book to scan and destroy. This was the basis of the 2025 legal settlement between Anthropic and authors.

In a statement to US media, Anthropic said Claude was trained on a mix of “publicly available web data, commercially acquired data sets, and data we generate ourselves”.

“Sourcing books is a widely used approach for training large language models across the AI industry,” it said. “None of our data acquisition programmes buy and destroy rare or antiquarian books.”

German booksellers are not convinced. They are sure, based on the vast orders, that the process is happening anyway via third-party service providers. And while Anthropic has been exposed in court, it is highly likely that many tech giants are engaged in similar practices.

And what is the end goal, after the AI has been trained? Berlin bookseller Zerfass thinks the consequences are vast – far more than a final death knell for booksellers like him.

“For me the intention here is to fragment human thinking even further,” he said, “to make people question even more their critical thinking abilities.”

German booksellers are convinced that large orders of old books are being used to feed large language models owned by AI giants. Photograph: Sean Gallup/Getty Images

In Galway, Kenny is less philosophical and more practical, yet equally alarmed at how AI is “terrifying our industry”.

“Booksellers tend to be at the forefront of a lot of moral dilemmas, and political and ethical issues. In most scenarios it is not for me to decide,” he said. “However, if they are taking a book and scanning it with a view to taking intellectual property, then Kennys do not want to be a part of that.”

联系我们 contact @ memedata.com