Canonical text (the exact bytes fingerprinted)

TRUVACE RECORD VERSION
record: TRV-2026-1562
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-10-16T10:41:38.303932Z
status: published
lens: trace
sector: education
headline: Oxford lets OpenAI train its AI models on Bodleian Library
dek: 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 wide…
gain_title: The Bodleian keeps the rights to the scans and plans to publish them openly online, which Oxford says will make rare historical material accessible to more people.
problem_title: Oxford's 2025 announcement did not say the digitised Bodleian texts would train OpenAI's models, and staff raised reputational and environmental concerns about the partnership.
trace_subject: Oxford Bodleian Library digitisation partnership with OpenAI
gain_reading: The Bodleian keeps the rights to the scans and plans to publish them openly online, which Oxford says will make rare historical material accessible to more people.
gain_evidence: The Bodleian keeps the rights to the scans and will begin publishing them openly online within months | make the material more accessible to a wider number for people
problem_reading: Oxford's 2025 announcement did not say the digitised Bodleian texts would train OpenAI's models, and staff raised reputational and environmental concerns about the partnership.
problem_evidence: the announcement did not state the material would be used for training OpenAI’s models | concerns from staff, including members of the Bodleian governance committee, about the reputational risk of partnering with OpenAI
quick_read: The University of Oxford has let OpenAI use texts digitised from the Bodleian Library to train its models. Internal documents say the scanned material was used to populate the company's training set. The partnership was announced in March 2025 as a way to digitise library content for students and researchers, without mention of model training.

The deal matters because AI developers are running short of fresh, high-quality data and are turning to historical book collections, which raises questions about disclosure and institutional trust. Oxford says the volume is modest and out of copyright and that it was open about the training element, but critics inside the university worried about reputation and energy use, and the scope of any wider digitisation remains unclear.
limitation: The extent of any wider digitisation is unclear, as the contract only raises the prospect of mass digitisation and Oxford disputes the characterisation of the project's scale and transparency.
tag: Dual reading
key_points: By June 2025, 125,000 images scanned from 19th- and 20th-century historical dissertations had been shared with OpenAI from the Bodleian collection. | Other scanned material includes about 10,000 16th-century broadside ballads, and staff have discussed Irish state papers, Marie Edgeworth's letters and Dorothy Hodgkin's penicillin notebooks. | Oxford is the only UK member of OpenAI's NextGenAI project, which also includes Boston Public Library, MIT and the University of Michigan. | Oxford says the Bodleian keeps its scans' rights, OpenAI's use is not exclusive, and the collections remain intact unlike books pulped after scanning elsewhere.
rundown: Oxford's spokesperson rejected the suggestion that the machine-learning element was hidden, saying digitisation was the primary interest but staff had been open that the project would also supply training data. The university describes the digitised amount as modest in scale and limited to out-of-copyright material.

The contract also raises the prospect of mass digitisation of the Bodleian's 14m items, though fragile and copyrighted collections would be exempt. Meeting minutes obtained through a freedom of information request also discussed an 'Ask the Bod' chatbot.

The article places the deal in a wider scramble for training data. Scraped websites are increasingly saturated with AI-generated material, and booksellers report orders for obscure titles that may be unlikely to exist online in digitised form. Anthropic has bought books and pulped them after scanning, though it says it does not buy and destroy rare and antiquarian books.
sources:
- journalism | The Guardian | https://www.theguardian.com/technology/2026/sep/26/oxford-university-bodleian-library-open-ai-chat-gpt | 2026-09-26
prev: 0000000000000000000000000000000000000000000000000000000000000000

sha256

646f1d86b2122f3ce74ca2e47224b6add3365628300766ac06f315df6d9bfbd1

previous

0000000000000000000000000000000000000000000000000000000000000000