Dark data is the info companies collect but don’t use—up to 55% of all data....

https://emilyscoolnews.urbanvellum.com/posts/what-is-garbage-in-garbage-out-for-ai-data-pipelines

Dark data is the info companies collect but don’t use—up to 55% of all data. This unused data still costs money to store and can create risks like security gaps. It also clutters AI projects, adding around 70% noise that makes results less accurate