Which Creative Commons licence should I use for research data?

The answer to which Creative Commons license you should use for research data is CC0, the public domain dedication, not CC BY or any license with conditions. CC0 was designed to remove copyright and database rights barriers that licenses with conditions leave in place. The decisive exception is that a CC license only applies where copyright or similar rights require permission. For a dataset, that is often the wrong question. Sui generis database rights, contractual access terms, and institutional policies are the real constraints. Check whether your funder or repository mandates a specific tool first. Then apply version 4.0 International unless a jurisdiction-specific port is already required by your institution.

Most researchers reach for CC BY because they want credit. CC BY version 4.0 requires attribution: the reuser must retain the creator's name, a copyright notice, a license reference, a warranty disclaimer, and a URI to the material. That works for a paper or a figure. For a structured dataset, attribution is a friction that blocks automated aggregation. It does nothing for the database rights that protect the selection and arrangement of facts.

Version 4.0 expressly addresses sui generis database rights. But a condition-based license still requires a downstream user to parse the condition for every field and every row. That is not how research data is used at scale.

What CC0 Waives

CC0 is a public domain dedication with a waiver, a fallback license where the waiver is not effective, and a non-assertion promise limited to remaining Copyright and Related Rights. It aims to remove copyright and related rights conditions as far as the affirmer is legally able. That includes the sui generis database right that protects the investment in collecting, verifying and presenting data. No condition-based Creative Commons license can do that, because the SA, NC and ND clauses operate only on top of copyright. If the database right is the primary right at stake, CC0 is the only tool that waives it.

Handling Credit Without A License Requirement

The cost is that CC0 has no attribution requirement. A reuser is not legally required to credit the source. Research integrity norms, community citation standards and funder policies may still demand credit. The official deed makes clear that credit is not a legal condition. If your institution or the data repository requires a citation trail, write the attribution expectation into the metadata and the terms of deposit rather than into the license itself.

A Creative Commons license operates on top of copyright. Exceptions and limitations such as fair use or fair dealing operate separately. If an exception authorises the use, the license's conditions do not apply to that use. For research data, the more common restriction is not copyright but a contractual access agreement or a repository's terms of use. Those are separate legal instruments. A CC license does not override them. Check your data sharing agreement and the repository's deposit terms before you choose a license.

Funder rules often specify a particular tool. The UK Research and Innovation policy on open access for data expects datasets to be made available with as few restrictions as possible. Many funding bodies point to CC0 as the recommended instrument for data. If your funder names a license, that is your answer. If your funder says nothing, your institutional policy may still require CC0 or CC BY for different categories of data. Separate the open license permission from the citation requirement. They serve different purposes and both should be addressed in the deposit metadata.

CC BY 4.0 is appropriate when the dataset includes substantial original expression: annotated images, narrative text, code documentation, or a curated commentary that is itself a copyrightable work. In that case, the copyright in the expression is the primary right, not the database right. CC BY gives you attribution for that expression without blocking reuse of the underlying facts.

Do not use CC BY for a bare numeric table, a gene sequence, a climate model output or a corpus of public records. Those are facts. Copyright does not protect facts.

CC BY-SA 4.0 requires that shared adaptations use BY-SA 4.0 or an officially compatible license. That condition is workable for prose or code documentation. For a dataset, it creates a practical chain of obligation that is almost impossible to verify. A reuser who merges two SA datasets must apply the SA condition to the merged product. The next reuser must do the same. The official compatibility list controls what licenses count as compatible, but the list is designed for creative works, not for data. Unless you have a specific legal reason to require share alike, avoid it for data.

A Creative Commons license and an academic citation serve different purposes. The license grants permission to reuse the work under copyright and related rights. The citation meets the scholarly norm of giving credit and allowing others to verify the source. Under version 4.0, attribution is a legal requirement only when the licensor supplies the required information. For CC0, there is no attribution requirement at all.

Write the citation expectation into your README file, the repository's metadata schema, and the terms of deposit. Do not try to enforce a citation through a license condition that was not designed for that purpose.

Verify Your Rights

Open the original source page of the official Creative Commons chooser tool. Use version 4.0 International for a new dataset unless a jurisdiction-specific port is required by your institution. Confirm that you own or control the rights to every component. Check your funder's open data policy and your repository's deposit terms.

Write The License Notice

Write the license notice in the metadata: identify the work and creator, name and link the exact license and version, link the source, and clearly exclude any material not covered. Do not assume a CC badge on a repository page resolves the conflict between the license and the platform's terms of use. Read the platform's current terms separately.

The single thing that most often goes wrong is that a researcher applies CC BY to a dataset that does not contain copyrightable expression. Then the attribution requirement blocks the automated reuse the dataset was created to enable. The dataset gets ignored by downstream tools. Or the researcher receives a complaint from a reuser who cannot comply with the attribution condition for thousands of records.

Apply CC0 for data. Write the citation expectation into the metadata. Use CC BY only when the dataset contains substantial original copyrightable work that you actually want to be attributed for.