Oxford lets OpenAI train models on Bodleian library texts

Staff and governance committee members raised concerns about reputational risks and the environmental impact of energy-intensive AI training. Oxford is the only UK member of OpenAI's NextGenAI project, which also includes US institutions like MIT and Caltech. An OpenAI spokesperson stated the collaboration helps preserve historical knowledge and ensures models reflect diverse cultures. The university maintains that the digitised material is out of copyright, modest in scale, and that the Bodleian retains rights to publish the scans openly online within months.

This move highlights a broader trend where AI developers seek fresh data from physical archives as scraped web content becomes saturated with AI-generated material. Unlike rival Anthropic, which has faced scrutiny for buying and pulping books for data, Oxford's deal keeps the physical collections intact. The university rejected claims that the training aspect was hidden, stating staff were aware the project would contribute training data.