The datasets behind DataDealRoom
1 dataset, sourced page by page and read again on a timetable, most recently 2026-10-08.
| Dataset | Headline figure | Sources | Read |
|---|---|---|---|
| AI data deal terms, buyer by buyer | 14 parties | 14 of read | 2026-10-08 |
From page to figure
A dataset is planned around one question and a named set of source pages. When a page is read, the exact sentence that carries the number is saved, along with where and when we found it.
Each candidate sentence passes through review, and any rejection carries a written reason. The median and range count one figure per source, one per breakdown where a source gives several. Sources without a figure are marked rather than guessed, and regular re-reads track what moves.
AI data deal terms, buyer by buyer
14 AI data buyers, brokers and data rooms were read for the deal terms they publish, each answer quoted from their own 19 pages, read 8 October 2026.
How the deal pays, by party: not stated 12 of 14, a licence fee or a one-off buyout 1 of 14, a revenue share on every sale 1 of 14 (14 parties, read 8 October 2026).