Wikimedia Commons

Commons:AI-generated media

This page is considered an official guideline on Wikimedia Commons.

It illustrates standards or behaviors which most editors agree with in principle and generally follow. Feel free to update the page as needed, but please use the discussion page to propose major changes.

Shortcut: COM:AI

AI-generated images of the Wikimedia Commons project logo

Media generated by artificial intelligence present unique challenges for licensing, attribution, and scope evaluation. As the technology is rapidly developing, and there are many legal and ethical gray areas concerning AI-generated media, best practices may change rapidly.

Are AI-generated media within the Commons project scope?

edit

Per the Commons project scope, only media that are realistically useful for an educational purpose should be hosted on Commons. Just because an AI image is interesting, pretty, or looks like a work of art, that doesn't mean that it is necessarily within the scope of Commons. While some AI-generated media fall within our scope, media that lack a realistic educational use may be nominated for deletion.

Can AI-generated media be lawfully hosted on Commons?

edit

Copyrights of person using generative AI

edit
 See also: Works Containing Material Generated by Artificial Intelligence.

In the United States, Indonesia[1] and most other jurisdictions,[2] only works by human authors qualify for copyright protection. In 2022 and 2023, the US Copyright Office repeatedly confirmed that this means that AI-created artworks that lack human authorship are ineligible for copyright.[3][4][5] The Commons community has rejected deletion requests that relied on such copyright claims, and tagged images generated by models such as DALL-E as {{PD-algorithm}}.

United Kingdom

edit

In the United Kingdom there is copyright protection for AI-generated media:

Unlike most other countries, the UK protects computer-generated works which do not have a human creator (s178 CDPA). The law designates the author of such a work as “the person by whom the arrangements necessary for the creation of the work are undertaken” (s9(3) CDPA). Protection lasts for 50 years from the date the work is made (s12(7) CDPA).[6]

In deletion requests, it may therefore be necessary to determine whether the work was generated in the United Kingdom, and if this can be affirmed, deletion may be necessary.

China (mainland)

edit

In China, AI-generated media may or may not be protected by copyright. In the specific case of Li v Liu (2023), a Chinese court ruled that an image generated by Stable Diffusion was protected by copyright. The court ruled that the image met the requirement of "intellectual achievement" due to the prompter's design of the prompt, selection of parameters, and selection of the image. It ruled that the image met the requirement of "originality" due to the image reflecting the prompter's aesthetic choices and personal judgments. The court's analysis was based on the prompter's complex prompt and process of adjustment and modification, and emphasized that the process was not merely "mechanical". Other cases of AI-generated media may not meet these requirements. It should also be noted that China uses a civil law system rather than a common law system, so other courts in China are free to decide differently.[7]

Hong Kong

edit

Similar to the United Kingdom section above, the Cap. 528 Copyright Ordinance provides a 50-year term for computer-generated works under 17. Duration of copyright in literary, dramatic, musical or artistic works:

“ (6) If the work is computer-generated the above provisions do not apply and copyright expires at the end of the period of 50 years from the end of the calendar year in which the work was made. ”

Copyrights of AI providers

edit

Although some AI providers may claim copyright ownership over generated works, such claims are unlikely to be legally valid or enforceable. This includes claims by AI providers that generated images cannot be used for commercial purposes (for example, Midjourney's free tier does not allow commercial use, but this is unlikely to be legally enforceable as Midjourney images are not copyrighted).

Copyrights of authors whose works were used to train the AI

edit
 
This AI-generated artwork is a derivative work of the original Mona Lisa.

Most image-generating AI models were trained using works that are protected by copyright. In some cases, such models can produce images that contain major copyrightable elements of those copyrighted training images, making these outputs derivative works. Accordingly, there is a risk that AI-generated art uploaded on Commons may violate the rights of the authors of the original works.

Special case: intentionally derivative works

edit
 See also: Commons:Derivative works.

When using AI to create media intended to imitate the style of a particular artist, such as using prompts that include the name of a specific artist, there is a risk that the output will be a derivative work of specific works by that artist. If those works are not yet in the public domain, it is possible that the output constitutes a copyright infringement. If clear similarities to specific copyrighted works are evident, those images should be nominated for deletion.

Special case: fan art

edit
 See also: Commons:Fan art.

Images that depict characters from or are based on proprietary works such as movies, TV shows, computer games, comic books or manga/anime might qualify as derivative works under copyright law. If the underlying work is copyrighted and not freely licensed, publication of such derivative works probably constitutes a copyright infringement. See the guideline on fan art for more specific guidance on how to distinguish infringing from non-infringing fan art.

edit

The United States Copyright Office has ruled that while individual images generated by Artificial Intelligence cannot be copyrighted, significant modifications by a human author can be. For example, in the webcomic Zarya of the Dawn, the images are generated by Midjourney (and are thus not copyrighted). However, the composition of the pages, the arrangement of the images, and the story are all copyrighted by a human.[5] When choosing whether to upload a piece of AI-generated media, make sure that it does not have sufficient human contributions that could be copyrighted.

Defamation and harassment

edit

Material that harasses, insults or defames a living person or a group of people should not be hosted on Commons under the Wikimedia Terms of Use and COM:DIGNITY. Questionable material should be discussed in a deletion request; in serious or repeated cases, the uploader or files in question should be reported to the administrators' noticeboard.

Terms of use of AI providers

edit

Per Commons:Non-copyright restrictions, "non-copyright related restrictions are not considered relevant to the freedom requirements of Commons". If a user violates the terms of use of an AI provider, that is a matter between them and the provider. The terms of use do not affect Commons, the rights of reusers, or the works themselves.

How should AI-generated media be handled?

edit

Attribution

edit

When giving attribution for a media file created by a prompt-based system such as a text-to-image model, best practice is to give the name of the specific AI engine used, followed by the name of the person who created the prompt. For example:

Author: Midjourney‎ AI; prompted by Jane Doe

If the file is a collaboration between a human and an AI (e.g. a human added original creative content other than the prompt), both should be listed as authors:

Author: Jane Doe and Midjourney‎ AI

Description

edit

Whenever you upload an AI-generated media file that was created by a prompt-based system, you are encouraged to document the prompt used to generate the media in the file description.

Licensing

edit

All AI-generated media should include the {{PD-algorithm}} licensing tag on the file page unless they have been substantially modified by a human or clearly derive from a specific non-AI-generated work. It is also recommended to include a free license tag to cover use in countries such as the United Kingdom and China (see above).

Categorization and templates

edit

Proper categorization of AI-generated media is important to make the files easier to find and, in particular, to distinguish from other images.

Discussions relating to AI-generated media

edit
 See also: Category:AI-generation related deletion requests.
edit

Art

edit

Fictional characters

edit

Countries where AI images are copyrighted

edit

Out of scope concerns with AI images

edit

Attribution concerns with AI images

edit

See also

edit

References

edit
  1. ↑ Can AI Images Be Copyrighted In Indonesia?. Mondaq (14 May 2024). "Despite being entirely new creations, images generated by AI cannot be copyrighted because they do not meet the criteria of uniqueness and personal expression required by the Indonesian Copyright Law."
  2. ↑ https://www.wipo.int/wipo_magazine/en/2017/05/article_0003.html
  3. ↑ https://www.smithsonianmag.com/smart-news/us-copyright-office-rules-ai-art-cant-be-copyrighted-180979808/
  4. ↑ https://www.copyright.gov/rulings-filings/review-board/docs/a-recent-entrance-to-paradise.pdf
  5. ↑ a b Zarya of the Dawn - US Copyright Office
  6. ↑ Artificial intelligence call for views: copyright and related rights. United Kingdom Intellectual Property Office, 23 March 2021. Retrieved 10 June 2023
  7. ↑ Chinese court declares that AI-generated image has copyright. TechnoLlama (9 December 2023).
Retrieved from "https://commons.wikimedia.org/w/index.php?title=Commons:AI-generated_media&oldid=958963521"