A report published Sept. 26, 2026, said internal University of Oxford documents described digitized material from the Bodleian Libraries as having been used to populate OpenAI’s training set. The same report said 125,000 images scanned from historical dissertations had been shared with OpenAI by June 2025.
What Oxford announced in March 2025
On March 4, 2025, the University of Oxford announced a five-year collaboration with OpenAI. Its public description centered on digitizing public-domain material from the Bodleian Libraries and making collections searchable and accessible to students and researchers. The pilot was described as covering 3,500 dissertations across disciplines, dated 1498–1884.
The reported material
The 125,000 figure counts images from historical dissertations, not dissertations themselves. The report also identified 10,000 sixteenth-century broadside ballads among scanned material. These are separate reported counts with different units.
Oxford’s account of the project
Oxford described the digitized material as modest in scale and out of copyright, and said OpenAI’s use was non-exclusive. The university also said the Bodleian retained the rights to make the scans available itself.
In a statement reported on Sept. 26, 2026, Oxford said the library would begin publishing the digitized material openly online within the next few months. The same report said Oxford staff had raised reputational concerns and concerns about the energy-intensive technology’s implications for the university’s environmental commitments.
