🔶 牛津大学允许 OpenAI 利用博德利图书馆资料训练 AI 模型

牛津大学已允许 ChatGPT 背后的公司利用其博德利图书馆的历史文献训练 AI 模型，科技公司正四处向学术机构搜寻新的数据来源。

根据内部文件，经由 OpenAI 数字化的博德利馆资料已被用于“充实 OpenAI 训练数据集”。

牛津大学于 2025 年 3 月宣布与该公司建立合作关系，使用 OpenAI 软件对这座世界知名大学图书馆中的文献进行数字化。校方当时表示，此举将使学生和研究人员能够更广泛地获取这些内容。

然而，该公告并未说明这些资料将被用于训练 OpenAI 的模型。模型通过被输入海量数据来识别文字中的模式，从而“学习”写出完整句子并执行其他认知任务。

OpenAI 一名发言人表示，公司“自豪地确保当今的 AI 模型为未来保存世界的历史知识”。

“每天有超过 10 亿人使用这项技术，因此，重要的是它能够反映不同的文化、历史和视角，”该发言人补充说。

通过信息自由请求获得的牛津大学会议纪要记录了教职员工的担忧，其中包括博德利馆治理委员会成员。他们担心与 OpenAI 合作的声誉风险，以及涉及高能耗技术的交易将对大学环境承诺造成影响。

书商也报告称，近期出现了一波对冷门书籍的订购潮，例如关于 18 世纪非洲农业器具的指南，或 20 世纪 50 年代赛车手的传记。二手书店店主推测，由于这些书目不太可能已在网上以数字化形式存在，它们代表了新的数据，可被下一代 AI 模型消耗。

被抓取的网站正越来越多地充斥 AI 生成材料，使其对于训练模型的价值下降；开发者因而转向实体书籍馆藏，其中往往包括历史文献。

OpenAI 已在名为 NextGenAI 的项目下，与波士顿公共图书馆、加州理工学院、麻省理工学院和密歇根大学等美国研究型图书馆达成类似协议。牛津大学是该项目唯一的英国成员。

截至 2025 年 6 月，博德利馆藏中已有 12.5 万张历史学位论文扫描图像被共享给 OpenAI，其中包括欧洲和美国大学在 19 世纪及 20 世纪撰写的博士论文。其他已扫描的文本包括一套罕见的 1 万份 16 世纪“单页民谣”收藏，其中包含歌曲歌词和乐谱；这些民谣曾在都铎时期的街角流传。工作人员还讨论了对 18 世纪爱尔兰国家文件、爱尔兰小说家玛丽亚·埃奇沃思的私人信件，以及多萝西·霍奇金的青霉素笔记进行数字化。

OpenAI 与牛津大学的合同还带来了对博德利馆 2300 万件馆藏进行大规模数字化的可能性。会议纪要还讨论了创建一个“Ask the Bod”聊天机器人的计划。

牛津大学一名发言人表示，正在数字化的文本数量“规模适中”，且仅涵盖已进入公版的材料。该发言人称，博德利馆保留扫描件的权利，并将在数月内开始将其公开发布到网上。

他们否认机器学习部分曾被对公众和学生隐瞒。该发言人表示，数字化是大学的首要关注点，但工作人员一直公开说明，该项目也将贡献训练数据。

根据这项协议，博德利馆的馆藏仍保持完整；这不同于其他地方收购二手书的做法——那些书籍在扫描后会被制成纸浆。OpenAI 的主要竞争对手 Anthropic 已花费数千万美元购入书籍，切除书脊以扫描内容，随后将书籍制成纸浆。Anthropic 表示，其不会购买和销毁珍本及古籍。

科技新闻网站 404 Media 也曾在一份二手书订单中放置追踪装置，并追踪到这些书被送往美国的一处亚马逊设施；在该处，书籍同样被拆解和扫描。

牛津大学发言人表示：“通过与 OpenAI 的项目数字化的材料规模适中、已进入公版，并且 OpenAI 对这些材料的使用并非独家。

“博德利图书馆同样保留自行提供这些数字化材料的权利。正如我们处理其他数字化合作项目产出一样，图书馆将在未来数月内开始在网上公开发布这些材料。

“事实上，该项目将让博德利馆能够使这些材料被更广泛的人群获取；否则，他们可能很难接触到这些材料。”

The University of Oxford has allowed the company behind ChatGPT to train its AI models on historical texts from its Bodleian Library, as tech companies scour academic institutions for fresh data.

The Bodleian material digitised by OpenAI has been used to “populate the OpenAI training set”, according to internal documents.

Oxford announced a partnership with the company in March 2025, using OpenAI software to digitise texts from the university’s world-famous library, which it said would make the content more widely available for students and researchers.

However, the announcement did not state the material would be used for training OpenAI’s models, which are trained to recognise patterns in words – and thus “learn” to write complete sentences and perform other cognitive tasks – by being fed vast amounts of data.

An OpenAI spokesperson said the company was “proud” to ensure “the AI models of today preserve the world’s historical knowledge for the future”.

“With more than a billion people using this technology in everyday life, it’s important it reflects different cultures, histories and perspectives,” they added.

Meeting minutes at the University of Oxford, obtained via a freedom of information request, record concerns from staff, including members of the Bodleian governance committee, about the reputational risk of partnering with OpenAI and the effect on the university’s environmental commitments of striking a deal involving an energy-intensive technology.

Booksellers have also reported a spate of orders for obscure titles such as a guide to agricultural implements in 18th-century Africa or biographies of 1950s car drivers. Secondhand bookshop owners have speculated that because the titles are unlikely to exist online in digitised form, they represent fresh data that can be consumed by the next generation of AI models.

Scraped websites are increasingly saturated with AI-generated material, making them less useful for training models, and developers have turned to physical, often historical, book collections.

OpenAI has struck similar agreements with US research libraries such as Boston Public Library, Caltech, MIT, and the University of Michigan under a project called NextGenAI. Oxford is the only UK member of the project.

By June 2025, 125,000 images scanned from historical dissertations had been shared with OpenAI from the Bodleian collection, including PhD theses from European and American universities written in the 19th and 20th centuries. Other texts scanned include a rare collection of 10,000 16th-century “broadside ballads” containing song lyrics and musical notes that were once circulated on Tudor street corners. Staff have also discussed digitising of 18th-century Irish state papers, the private letters of Irish novelist Marie Edgeworth, and Dorothy Hodgkin’s penicillin notebooks.

The OpenAI contract with Oxford also raises the prospect of the mass digitisation of the Bodleian’s collection consisting of 23m items. The minutes also discussed the creation of an “Ask the Bod” chatbot.

A spokesperson for the University of Oxford said the amount of text being digitised was “modest in scale” and covered only out-of-copyright material. The Bodleian keeps the rights to the scans and will begin publishing them openly online within months, the spokesperson said.

skip past newsletter promotion

Sign up to TechScape

A weekly dive in to how technology is shaping our lives

after newsletter promotion

They rejected the suggestion that the machine-learning element had been hidden from the public and students. Digitisation was the university’s primary interest, said the spokesperson, but staff had been open that the project would also contribute training data.

The Bodleian’s collections remain intact under the deal, unlike with secondhand book acquisitions elsewhere that are being pulped after scanning. Anthropic, OpenAI’s close rival, has spent tens of millions of dollars acquiring books and slicing off their spines so their contents can be scanned before having them pulped. Anthropic has said it does not buy and destroy rare and antiquarian books.

A tech news site, 404 Media, also placed a tracking device inside a secondhand book order and traced it to an Amazon facility in the US, where the books were also dismantled and scanned.

The Oxford spokesperson said: “The material digitised through the project with OpenAI is modest in scale, out of copyright, and OpenAI’s use of the material is not exclusive.

“The Bodleian libraries also retain the rights to make the digitised material available themselves, and the library will begin to publish these materials openly online in the next few months, as we do with outputs from other digitisation partnerships.

“The project will in fact allow the Bodleian to make the material more accessible to a wider number for people, who might otherwise have found it difficult to access.”