Z-Library是互联网上最大的海盗电子书库之一，免费提供了上百万本书籍和论文。
对负担不起昂贵的教科书和期刊订阅费的学生和研究人员来说，它可以成为有用的资源。 与此同时，出版商花费了多年时间试图将其取下，并附有取走通知和封杀命令，而美国当局则起诉其指称的运营商。
尼日利亚大学的两名图书管理员撰写的《学术图书馆新评论》杂志上
主要作者与TorrentFreak分享他们作品的副本。 虽然这些结论并不立即成为新闻，但有一个提法现在非常突出。
种子碎片文章不存在
在关于强制执行Z-Library的一节中，提交人写道，包括Elsevier和Springer Nature在内的主要学术出版社采取了类似的法律行动。 参考文献为2021"TorrentFreak"一文，如下所示。
问题是，这种特殊的法律诉讼对我们来说是新的。 参考部分将文章列为"Publishers Congress on shadow library"并链接到TorrentFreak主页。 然而，我们从未发表过一篇有这一标题的文章。
我们确实报告说，Elsevier和Springer Nature获得了针对Sci-Hub的英国封杀令，几个Z-Library域因版权被侵犯被暂时中止了声称。 然而，我们不知道针对Z-Library的这一特别法律行动。
更仔细地看，我们看到更多的声称似乎不适合。 例如，该报对Z-Library刑事案件的叙述指出，美国司法部于2022年 " 逮捕了个人 "。
司法部自己的新闻稿称声称，当时我们所报道的两名操作人员“应美国的要求”在阿根廷被捕。 报纸为这份新闻稿提供的标题也不存在。
可用但不准确
这些不存在的引述似乎是合理的，因为它们以适当的背景列举了真正的出版物和组织。 但这并不表示他们准确。 当我们查阅全部参考清单时，我们注意到这适用于更多的来源。
论文引用了2019年哈佛法律与技术杂志由IP法学教授Daniel Gervais撰写的一篇文章，题为“影子图书馆阴影下的知识产权的未来 ”。 所引用的题材表没有包括
由卡罗尔·特诺皮尔和同僚撰写的一篇2019 College & Research Library的文章也是如此。 引用的问题没有特诺皮尔的文章。 她的确共同撰写了一篇关于类似主题的2019年的文章，但该文章出现在了不同的期刊上，标题不同。
其他无法找到的参考文献包括2019年的一篇属于Björn Brembs的文章和名为Smith、Piper、Okike和Aronson的作者的文章，以及关于“影子图书馆的隐蔽危险”的Kaspersky报告。
我们无法找到本文44个参考文献中的至少10个
百分之百的人工智能？
显而易见的问题是这些神秘的参考来源。 该文件的承认可能为正确方向提供了提示，因为它们提到在研究和写作过程中使用了AI。
撰文者写道，“我们确认共识AI被用于通过大量文献筛选和提取相关信息”。 他们补充说，ChatGPT被用来改进部分纸张的语言。
AI聊天机是发明参考文献的已知工具，经常有可信的标题被记入该领域真正的专家名下。 这可以解释调查结果
也透过AI检测工具Pangram，对于这两节，该工具报告说，“100%的文本可能是人工智能生成的。” 这些类型的工具并不完美，可以产生虚假的阳性，但符合画面。
如果AI造成这些错误，那将是相当讽刺的。 这意味着泰勒和弗朗西斯出版的一本商业，同行评审的期刊中的一篇论文依赖于AI写作"Z-Library". 同时，许多权利持有者正在起诉AI公司使用影子图书馆来培训AI模型。
Z-Library等网站会损害出版商的收入
“广泛使用Z图书馆对学术出版的可持续性构成重大挑战。 出版商依靠订阅费和付费墙来资助同行评审、编辑和分发学术著作，“该书的原文是，据推测也是由大赦国际制作的。
本文并未提及同行评审一般不为其工作付费
出版商调查
主要作者起初与我们分享了一篇论文，也问Gervais、Brembs和Tenopir是否写了报章
称表示正在调查此事。
“我可以证实，这已提交给我们的编辑和出版道德与廉正小组，他们将调查你所关心的融资问题。”
出版商在调查期间不会再发表进一步的评论。 该文件是否得到纠正或撤回，尚待观察。
—
本文载有“Z-图书馆和全文查阅：知识民主化还是损害版权道德？”
本文由 AI宝库自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。
Z-Library is one of the largest pirate ebook libraries on the Internet, offering millions of books and papers for free.
For students and researchers who can’t afford expensive textbooks and journal subscriptions, it can be a useful resource. At the same time, publishers have spent years trying to take it down, with takedown notices and blocking orders, while U.S. authorities indicted its alleged operators.
This tension is the subject of a paper published earlier this month in the journal New Review of Academic Librarianship, written by two librarians at Nigerian universities.
The peer-reviewed academic paper doesn’t pick a side and the lead author kindly shared a copy of their work with TorrentFreak. While the conclusions are not immediately newsworthy, there was one reference that stood out right away.
A TorrentFreak Article That Doesn’t Exist
In a section on enforcement against Z-Library, the authors write that major academic publishers, including Elsevier and Springer Nature, have taken similar legal action. The reference is a 2021 TorrentFreak article, as shown below.
The problem is that this particular legal action is new to us. The reference section lists the article as “Publishers continue crackdown on shadow libraries” and links to the TorrentFreak homepage. However, we have never published an article with that title.
We did report that Elsevier and Springer Nature obtained a UK blocking order against Sci-Hub, and that several Z-Library domains were temporarily suspended over copyright infringement claims. However, we’re not aware of this particular legal action against Z-Library.
Looking more closely, we see more claims that don’t seem to fit. The paper’s account of the Z-Library criminal case, for example, states that the U.S. Department of Justice “arrested individuals” in 2022.
The DOJ’s own press release says that the two alleged operators, whose indictment we covered at the time, were arrested in Argentina “at the request of the United States.” The title the paper gives for that press release doesn’t exist either.
Plausible but Inaccurate
These non-existent citations appear plausible, as they cite real publications and organizations in a fitting context. However, that doesn’t mean they are accurate. When we went over the full reference list, we noticed that this applies to more sources.
The paper cites a 2019 Harvard Journal of Law & Technology article by IP law professor Daniel Gervais, titled “The future of intellectual property in the shadow of shadow libraries.” The table of contents of the cited issue doesn’t include it, and the cited pages are taken up by two articles on other topics.
The same applies to a 2019 College & Research Libraries article attributed to Carol Tenopir and colleagues. The cited issue has no article by Tenopir. She did co-author a 2019 article on a similar topic, but it appeared in a different journal, under a different title.
Other references that couldn’t be located include a 2019 article credited to Björn Brembs and articles credited to authors named Smith, Piper, Okike and Aronson, as well as a Kaspersky report on “the hidden dangers of shadow libraries.”
In total, we had trouble finding at least ten of the 44 references in the paper, as well as others that didn’t necessarily back the stated claims.
100% AI?
The obvious question is where these mystery references came from. The paper’s acknowledgements may offer a hint in the right direction, as they mention that AI was used in the research and writing process.
“We confirm that Consensus AI was used in sifting through large volume of literature and extract relevant information,” the authors write. They add that ChatGPT was used to improve the language of parts of the paper.
AI chatbots are known to invent references, often with plausible titles credited to real experts in the field. This could explain the findings, although it doesn’t prove that AI wrote these references.
We also ran the paper’s abstract and conclusion through the AI detection tool Pangram. For both sections, the tool reported that “100% of the text is likely AI-generated.” These types of tools are not perfect and can produce false positives, but it would fit the picture.
It would be rather ironic if AI caused these errors. It would mean that a paper in a commercial, peer-reviewed journal, published by Taylor & Francis, relied on AI to write about Z-Library. Meanwhile, many rightsholders are suing AI companies for using shadow libraries to train AI models.
Interestingly, the paper explicitly warns that sites such as Z-Library hurt publishers’ revenue, which funds peer review and editing.
“The widespread use of Z-Library poses significant challenges to the sustainability of academic publishing. Publishers rely on subscription fees and paywalls to fund the peer review, editing, and distribution of scholarly works,” it reads, presumably generated by AI as well.
The paper doesn’t mention that peer reviewers generally aren’t paid for their work, an issue that was central to a recent lawsuit against major academic publishers, including Taylor & Francis.
Publisher Investigates
The lead author initially shared a copy of the paper with us, but didn’t answer our follow-up questions about the mysterious references. We also asked Gervais, Brembs and Tenopir whether they wrote the papers that are credited to them, but they didn’t respond either.
Publisher Taylor & Francis did reply and says that it’s looking into the matter.
“I can confirm that this has been referred to our Editorial and Publishing Ethics & Integrity teams, who will investigate the concerns you have raised,” a Taylor & Francis spokesperson informed TorrentFreak.
The publisher will not comment further while the investigation is ongoing. Whether the paper will be corrected or retracted has yet to be seen.
—
The paper “Z-Library and Full-Text Access: Democratizing Knowledge or Undermining Copyright Ethics?” is available here.
Auto-fetched and translated by AI Gem Box. All rights belong to the original author.