AI Data Provenance – Historical Shorthand

Our CEO, David Rogers gave a talk at the Techworks AI event at Bletchley Park in early 2025. The talk goes into detail about Copper Horse’s research into 17th century shorthand. This started as a ‘toy AI model’ to exercise the TAIBOM project, to also be used for security testing. The model turned out to be very useful to all parts of the project. It opened up a whole new area of discussion around the provenance of archival information and how it is protected. Future AI models will be much better if the datasets themselves are of high quality, but to do this, they need to be digitally signed and attributed to the institutions and researchers who have created them. AI models in general will benefit from tamper-resistance in all aspects whether it be the model itself, the training data or inference engines and weights.

The talk gives a small insight to the tumulous events of the 17th century, a period which saw the meteoric rise of shorthand, particularly in England during years of civil war, dissenting religious groups and many years of scientific revolution. Copper Horse continues to work on the models outlined here, expanding its datasets and building on the user interface to make it usable for historical researchers. Watch the talk here: