Dataset licenses explained
Most dataset licensing questions come down to three families: Creative Commons licenses such as CC-BY and CC0, database-specific licenses such as ODbL, and research-only or non-commercial terms attached by the publisher. This is general information, not legal advice — read on for what each family actually permits and where to check before you use a licensed dataset.
The short answer
CC-BY lets you use, adapt and redistribute a dataset, including commercially, as long as you credit the creator. CC0 removes even the attribution requirement, placing the work in the public domain. ODbL is built specifically for databases and requires attribution plus share-alike for any redistributed adaptation. Research-only or non-commercial terms restrict use to academic or non-profit contexts and have to be read individually, since their exact boundaries vary by publisher. This page explains the mechanics; it is general information, not legal advice.
Creative Commons licenses: CC-BY and its family
Creative Commons publishes a standard set of licenses that combine four conditions — attribution, share-alike, non-commercial and no-derivatives — into six named licenses, listed in full on the Creative Commons license list. The most common one you will see on datasets is CC BY, which allows you to distribute, remix, adapt and build on the material in any medium, including commercially, as long as you give credit to the creator in the way they request. Variants add further conditions: CC BY-SA requires that anything you build from the data be shared under the same license; CC BY-NC restricts use to non-commercial purposes; CC BY-ND prohibits distributing modified versions. Before using a Creative Commons–licensed dataset, check which specific variant applies — the letters after "BY" change what you are allowed to do.
CC0: no conditions at all
CC0 1.0 Universal is not a license in the traditional sense — it is a public domain dedication. A creator who applies CC0 waives essentially all copyright and related rights, worldwide, to the fullest extent the law allows. Under CC0 you can copy, modify, distribute and use the work, even commercially, without asking permission and without giving attribution. It is the most permissive option a dataset publisher can choose, and it is common on government and reference datasets where the publisher wants zero friction to reuse.
ODbL: built specifically for databases
The Open Database License (ODbL), published by Open Data Commons, addresses a gap that Creative Commons licenses were not originally designed for: many jurisdictions grant separate legal protection to the structure of a database on top of copyright in its individual contents. ODbL requires that you attribute the source, and that if you publicly use an adapted version of the database, you offer that adapted version under ODbL as well — a share-alike condition applied specifically to the database, not just to any creative content inside it. It is important to note that ODbL governs rights over the database itself; if the individual records also carry separate copyright, the publisher typically pairs ODbL with a content license such as CC BY-SA. OpenStreetMap is the best-known dataset licensed this way.
Research-only and non-commercial terms
Not every dataset uses a standard open license. Many academic and government-adjacent datasets are released under custom "research-only," "academic use," or non-commercial terms defined directly by the publisher rather than by Creative Commons or Open Data Commons. These typically permit use for study, benchmarking and non-profit research, but explicitly prohibit incorporating the data into a commercial product, and often prohibit redistributing the raw files outside the research context. Because these terms are custom, the exact line between permitted and prohibited use is set by the specific license text, not by convention — always read the publisher's own terms page rather than assuming it matches another dataset's non-commercial clause.
License comparison
| License | Commercial use | Attribution required | Share-alike required |
|---|---|---|---|
| CC BY | Allowed | Yes | No |
| CC0 | Allowed | No | No |
| ODbL | Allowed | Yes | Yes, for adapted databases |
| Research-only / non-commercial | Prohibited | Usually yes | Varies by publisher |
FAQ
Can I use a CC-BY dataset for a commercial product?
Yes. CC BY allows commercial use as long as you credit the original creator in the way the license or the publisher's instructions specify. You do not need permission and you do not owe royalties, but you do need to keep the attribution intact wherever the data or work derived from it is used.
What is the difference between CC0 and CC-BY?
CC0 is a public domain dedication: the creator waives essentially all rights, and you can use the data with no conditions at all, not even attribution. CC BY keeps copyright in place but grants a broad license to use, adapt and redistribute the work, provided you credit the creator. If a project needs zero-friction reuse, CC0 is the more permissive choice; if the publisher wants credit, CC BY is the more common one.
What does a research-only license actually restrict?
A research-only or non-commercial term typically permits use for academic study, benchmarking and non-profit research, but prohibits use in a commercial product or service, and often prohibits redistribution of the raw data outside the research context. The exact boundary of what counts as commercial use is defined by the license text itself, not by a general convention, so it has to be read directly rather than assumed.
This is general information, not legal advice. If a license question affects a real deal or product launch, confirm the specific terms with qualified counsel.
Not sure a dataset's license fits your use case?
DeepSData can run a real availability search and verify sources for you — checking license terms, provenance and coverage against the requirements you give us, so you are not left guessing what a custom research-only clause actually permits.