Home / Education / Oxford Library Texts Fuel AI Data Race
Oxford Library Texts Fuel AI Data Race
26 Sep
Summary
- OpenAI is training AI models using digitized historical texts from Oxford's Bodleian Library.
- Staff expressed concerns about reputational and environmental risks from the partnership.
- Other libraries and secondhand bookshops are also supplying data to AI companies.
The University of Oxford has entered into an agreement with OpenAI, the creator of ChatGPT, permitting the company to utilize digitized historical texts from its renowned Bodleian Library for AI model training. Internal documents reveal that this material has been incorporated into OpenAI's datasets, a fact not explicitly detailed when the partnership was announced in March 2025.
Staff members at Oxford have voiced significant concerns, citing potential reputational damage and environmental impacts associated with the energy-intensive nature of AI technology. These worries emerged from meeting minutes obtained through a freedom of information request. The trend highlights a broader push by tech companies towards physical archives as online data sources become saturated with AI-generated content.
OpenAI's 'NextGenAI' project, which includes Oxford as its sole UK participant, has similar agreements with U.S. institutions like Boston Public Library and MIT. By June 2025, over 125,000 scanned images from historical dissertations were shared with OpenAI from the Bodleian collection. The university spokesperson emphasized that the digitised material is modest in scale, out of copyright, and that the Bodleian retains rights to publish the scans online within months.
