Harvard adds copyright-free fuel to the AI fire.

With funding from Microsoft and OpenAI, the university’s Institutional Data Initiative (IDI) is releasing a dataset for training AI that contains nearly one million public-domain books — around five times larger than the controversial Books3 dataset.

The aim is to “level the playing field” for smaller AI developers who don’t have access to the massive datasets used by tech giants, according to IDI executive director Greg Leppert.