Story perspectives

Harvard, Microsoft, OpenAI Release Million-Book AI Dataset

11/9/2025

28 6

Summary

Story summary

Harvard University, with Microsoft and OpenAI, released nearly one million public domain books dating to the 15th century to aid AI researchers.
The collection comprises 394 million scanned pages and offers greater linguistic diversity than typical AI datasets.
The project aims to provide a more reliable data source for AI training than previously used online content.
However, concerns about outdated and harmful content persist, prompting efforts to guide responsible AI usage.

1 of 1

Story summary

Harvard University, with Microsoft and OpenAI, released nearly one million public domain books dating to the 15th century to aid AI researchers.
The collection comprises 394 million scanned pages and offers greater linguistic diversity than typical AI datasets.
The project aims to provide a more reliable data source for AI training than previously used online content.
However, concerns about outdated and harmful content persist, prompting efforts to guide responsible AI usage.