A common assumption is that data labeled "open" can be used for anything. An open data license does grant permission in advance, so you do not have to ask, but almost every such license attaches conditions, and the conditions differ in ways that matter for a data project: whether you must credit the source, whether anything you build from it must be shared under the same terms, and whether commercial use is allowed. This explainer covers what an open data license is, the main license families you will meet, how they relate to terms of service and privacy law, and the mistakes that most often turn permitted use into a problem.
What an open data license is
An open data license is a standard set of terms under which a rights holder lets anyone use, copy and share a dataset or content, subject to stated conditions. It works on top of the rights the publisher holds, mainly copyright and, in the European Union and the United Kingdom, the database right. Without a license, those rights would require you to ask permission for many uses; with one, permission is already given as long as you meet the conditions.
Two limits follow from that. A license can only grant rights the licensor actually holds, so it does not cover third-party material that happens to appear in the dataset. And it governs copying and reuse, not everything else: it does not make personal data free of privacy rules, and it does not grant access to systems.
Creative Commons licenses and their building blocks
Creative Commons licenses are built from four conditions, combined into six licenses:
- BY (attribution): credit the creator, link or name the license, and indicate if changes were made. Every Creative Commons license includes it.
- SA (share-alike): if you share an adapted version, you must license it under the same or a compatible license.
- NC (non-commercial): use is permitted only for purposes not primarily aimed at commercial advantage or monetary compensation.
- ND (no derivatives): you may share the material only unchanged.
Separately, CC0 is a public domain dedication: the rights holder waives their rights as far as the law allows, with a fallback license where a waiver is not possible. For data, CC0 and CC BY are the most practical, because they set the fewest conditions on combining and redistributing.
The version number matters. The 4.0 licenses explicitly cover the database right as well as copyright, while earlier versions handled it differently depending on the version and the country adaptation. When a dataset says "CC BY" without a version, check the license link to see which one applies.
Open Data Commons licenses
A second family was written specifically for databases. It has three main licenses:
- PDDL, a public domain dedication similar in effect to CC0.
- ODC-BY, which requires attribution, similar in spirit to CC BY.
- ODbL, which adds share-alike for databases: if you publicly use an adapted version of the database, you must offer that adapted database under the ODbL or a compatible license. Works produced from the database, such as a chart or a map, are not themselves subject to share-alike, but they must carry a notice saying that the database is used under the ODbL.
The distinction between the database and works produced from it is the one to get right. Publishing a report with figures drawn from an ODbL dataset is usually straightforward; publishing your enriched copy of the dataset brings the share-alike obligation with it.
Suppose, hypothetically, a team combines an ODbL dataset of public transit stops with its own collected list of nearby shops. An internal dashboard that shows the combined data to employees raises a different question from a public download of the merged table. The public download is the case where the merged table could have to be offered under the ODbL, so the team should decide before merging whether its own list can be shared that way, and check how the license treats the combination if it cannot.
Government licenses
Many governments publish data under their own licenses, usually attribution-based, such as the Open Government Licence used for much public sector information in the United Kingdom. Some national portals use Creative Commons licenses instead. In the United States, works created by federal government employees as part of their duties are generally not protected by copyright, although state and local governments set their own policies, and federal sites can include material from contractors or third parties that carries its own terms.
Government licenses often exclude certain items, such as personal data, logos and material owned by others. The exclusions are usually listed in the license itself, and they are worth reading before collecting a whole portal.
Where an open license stops
A license covers reuse, not access. A CC BY dataset can sit on a site whose terms ask you to use the bulk download or API rather than crawl the pages, and robots.txt still applies to any crawling, as the guide on how to read a robots.txt file explains. When a source offers an official download under an open license, collecting it that way is usually better for everyone than scraping the pages that display the same records: the source carries less load, and the license tells you clearly what you received.
Common mistakes with licensed data
Treating non-commercial as obvious. Whether a given use is primarily commercial is often unclear, for example internal analysis at a company, or a research project funded by a business. If the project has any commercial side, treat NC data as needing a closer look rather than assuming it is fine.
Losing attribution in the pipeline. Attribution is easy to give in a one-off report and easy to lose in a dataset that merges dozens of sources. Store the source, the license and its version as fields on every record or batch, so credits can be produced whenever the data is published. Whether collection runs in-house or through a provider such as ScrapingIsNotACrime, that provenance should travel with the data.
Combining share-alike licenses. Two share-alike licenses can conflict, since each requires the combined work to carry its own terms. Some licenses publish compatibility statements; without one, combining them in a published dataset needs a careful look.
Assuming no license means no restrictions. Material with no stated license is not open by default. The publisher's rights still apply, and the absence of a license is a reason to check the terms or ask.
Assuming the site's license covers everything on it. User comments, photos from agencies and embedded content often belong to others. When the question is whether a specific planned use of a specific dataset is allowed, especially for publication or resale, a lawyer familiar with intellectual property in the relevant jurisdiction can give an answer that a general explainer cannot.