The University of Oxford has permitted OpenAI to utilize digitized historical materials from its Bodleian Library to "populate the OpenAI training set," according to internal university documents. This arrangement is part of a five-year collaboration between Oxford and OpenAI, which began in March 2025. The partnership aims to digitize public domain materials from the Bodleian collections, making them more accessible to students and researchers globally.
The initial announcement of the collaboration in March 2025 by Oxford University did not explicitly state that the digitized content would be used for training OpenAI's artificial intelligence models. However, an Oxford spokesperson has disputed claims that this aspect of the project was concealed from the public or university staff, stating that the university's primary interest was digitization and that staff were open about the project also providing training data.
Internal meeting minutes from Oxford University, obtained through a freedom of information request, indicate that some staff members, including those on the Bodleian governance committee, expressed concerns. These concerns centered on the potential reputational risk associated with partnering with OpenAI and the implications for the university's environmental commitments, given the energy demands of AI technologies.
The pilot phase of the digitization project, which commenced in February 2025, focuses on 3,500 public-domain dissertations dating from 1498 to 1884. By June 2025, approximately 125,000 images from historical dissertations had been shared with OpenAI from the Bodleian collection. Other materials scanned include a collection of 10,000 16th-century "broadside ballads." An Oxford spokesperson confirmed that the material being digitized is "modest in scale," out of copyright, and that OpenAI's use is not exclusive. The Bodleian Libraries retain the rights to the scans and plan to publish these materials openly online within months.
OpenAI has stated that using historical texts for training helps ensure that AI models reflect diverse cultures, histories, and perspectives, which is important given the technology's widespread use. The collaboration also includes research grants, computing resources, and access to OpenAI models like o1 and 4o for Oxford researchers through OpenAI's NextGenAI initiative, which has committed $50 million in grants and resources to participating institutions. Oxford is the sole UK institution participating in the NextGenAI project, which also includes US research libraries such as Boston Public Library, Caltech, MIT, and the University of Michigan. The partnership also involves expanding Oxford's use of ChatGPT Edu.