9Publish with a visible free licence

Works are protected by copyright from the moment they are created, without needing to register them or to say so. The consequence is that a resource published without a licence has all rights reserved, so other people can view it, but cannot copy, adapt or share it safely, even if its author has no objection.

The licence is what turns a published resource into a shared resource. The Recommendation on Open Educational Resources (OER) of the United Nations Educational, Scientific and Cultural Organization (UNESCO), adopted in 2019, defines them as materials that reside in the public domain or have been released under an open licence that permits no-cost access, re-use, re-purposing, adaptation and redistribution by others. Therefore, a resource without an open licence is not an open educational resource, even if it can be freely consulted.

A resource created with vibe coding brings together two types of work, which use different licences. The content is the texts, questions, images and explanations, and is usually published under a Creative Commons (CC) licence. The code is what makes the material work, and it uses licences designed for computer programs.

Not all Creative Commons licences are free, in the sense set out in the Definition of Free Cultural Works: a free work can be studied, copied, modified and redistributed by anyone and for any purpose. Creative Commons itself uses this definition to distinguish its free licences, CC BY and CC BY-SA, from those that are not. Each condition is indicated by two letters. BY requires authorship to be credited, and SA (ShareAlike) requires the same licence to be kept in adaptations. Those that include the NC (NonCommercial) condition, which prohibits commercial uses, or the ND (NoDerivatives) condition, which prohibits adaptations, limit precisely what one wants to allow. For content, therefore, two free licences remain:

• CC BY. It allows any use, including adapting the material, with the sole condition of crediting the author. It gives the widest dissemination, although improved versions may be published under a more restrictive licence.
• CC BY-SA. It adds the obligation to publish adaptations under the same licence, so that improvements remain free for the whole educational community.

It is advisable to avoid the NC condition, even though it may seem the most prudent, because it is not always clear what a commercial use is. Creative Commons itself states that it depends on the circumstances and the intention with which the material is used, and not on the type of organisation, and that it cannot clarify which uses are commercial. The use by a private school or by a platform with advertising, for example, is debatable.

If the aim is that nobody appropriates the material, that protection is given by the SA condition. A company may use a resource under CC BY-SA, but it has to publish its adaptations under the same licence, so that the improvements return to the community. The NC condition, on the other hand, closes two doors. The material is left out of repositories that only accept free resources, and it cannot be mixed with other free materials. According to the Creative Commons FAQ, an adaptation of SA material has to be published under the same licence, so a text or an image under CC BY-SA cannot be incorporated into a resource under CC BY-NC-SA.

For code, the most widely used free software licences are four:

• MIT. The simplest, which allows any use provided the authorship and licence notice is kept.
• Apache 2.0. Equally permissive, but it asks for changes to be indicated and includes a patent grant.
• MPL 2.0 (Mozilla Public License). It requires modifications to files under this licence to be shared, even if they are combined with closed code.
• AGPL v3 (Affero General Public License). It requires the code of modified versions to be shared under the same licence, also when they are offered as a service on a web page and not distributed. That last condition takes effect when the program runs on a server, since a resource that runs entirely in the browser hands its code to the person who opens it.

This guide recommends CC BY-SA for content and AGPL v3 for code, because it ensures that improvements return to the educational community. For maximum adoption, CC BY with MIT or Apache 2.0 can be chosen. The licence generator prepares the notice for any of these combinations, as text or as HTML ready to paste. The article «Licencias libres para docentes» (Free licences for teachers, in Spanish) goes deeper into the subject, with comparative tables, practical cases and answers to the most common questions.

The licence must be visible to fulfil its function. The minimum is for the material itself to state, in a visible place such as the footer, the authorship, the name of the licence and the link to its official text. That way the licence accompanies the material when it is downloaded or embedded in another page.

When the material is a project with several files, it is also advisable to add the file with the full text of the code licence and to mention both licences in the documentation. In a repository, the LICENSE file contains the licence of the code and another one, for example LICENSE-CONTENIDOS, that of the content. At the top of each code file a line such as SPDX-License-Identifier: AGPL-3.0-or-later is also advisable, which identifies the licence in a standardised way. All this can be requested from the artificial intelligence (AI) from the start, as the instructions file for the AI does, so that the material is generated with its licence already in place.

The question of who owns what the AI generates has two parts. The first is whether the company offering the AI claims any rights, and the answer is no. The terms of the main services, consulted in September 2026, agree on this point. OpenAI states that the user owns the output and assigns to them all the rights it might have over it. Anthropic likewise assigns to the user its rights over the outputs. Google states that it will not claim ownership of original content generated with its services. The material, therefore, can be published, shared and licensed without the company having any objection.

The second part is whether there is a copyright over what is generated that can be enforced against other people, and that is not decided by the terms of use, but by the law of each country. The companies themselves assign their rights «if any», because nobody can assign what they do not have. In Spain, the Intellectual Property Law considers the author to be the natural person who creates a work, and the question is whether the person giving instructions to the AI creates the work.

There are two positions on that question. The first holds that AI is a tool, like the brush or the camera, and that authorship belongs to the person who wields it. It is the one taken by British law since 1988, which attributes the authorship of a computer-generated work to the person who makes the arrangements necessary for its creation, and the one followed in 2023 by the Beijing Internet Court when it recognised copyright in an image generated with AI, because of the decisions the user took in choosing and adjusting the instructions. The second position holds that authorship requires the person’s creative decisions to be perceptible in the result. It is that of the United States Copyright Office, which concluded in 2025 that protection does not extend to the mere formulation of instructions, but does extend to what the person has created and can be recognised in the result, and to the creative arrangements and modifications they make to it. In Spain there is still no ruling, but the first European courts that have ruled follow the second position. In 2026, the Higher Regional Court of Düsseldorf, in line with two other German courts, held that an image generated with AI is only a work when it reflects specific creative decisions of the person, and that general instructions, however many, or choosing among several proposals, are not enough. The Court of Justice of the European Union has not yet ruled, so the question is not settled.

This guide does not need to resolve it, because its recommendation is the same under either position: publish the material with a free licence and state that it was created with AI, as recommendation 8 asks. Under the first position, the licence covers the whole material. Under the second, it covers the person’s contribution, and the rest would be left without rights that prevent its reuse. In both cases the material can be shared. It should be borne in mind that, under the second position, conditions such as share-alike could only be enforced on the human contribution. In educational vibe coding that contribution is extensive, since the person decides what is built, provides the content and corrects the result over many sessions, and recommendations 7 and 8 serve to record it.