Guides · Updated 2026-08-23
A dataset without a usable license is a dataset you cannot build on. Here is the real license picture across the 1000 most downloaded LeRobot datasets, and what each common choice means.
| License | Datasets | Share |
|---|---|---|
| apache-2.0 | 591 | 59% |
| cc-by-4.0 | 170 | 17% |
| none | 135 | 14% |
| mit | 38 | 4% |
| other | 37 | 4% |
| cc-by-nc-4.0 | 11 | 1% |
Apache-2.0 and MIT: use it, modify it, sell products built on it, just keep the notice. The friendliest choices. CC-BY-4.0: same freedom, credit the author. Anything with NC (like cc-by-nc-4.0): non-commercial only, companies cannot train sellable models on it. "None" or "other": legally you have no permission at all. 172 of 1000 datasets sit in that last bucket.
Pick the license on upload, not later. Datasets with no license get flagged in Datum's checks because buyers filter them straight out. If you want your data inside the next open robot model, Apache-2.0 or CC-BY-4.0 is the answer.
Filter for commercial-use in the finder. It reads the declared license on every dataset and marks which ones are safe for commercial training.