Commons:AI images of identifiable people

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.

When dealing with photographs of people, we are required to consider the legal and moral rights of the subject. Images of people created by generative artificial intelligence (AI) raise similar legal and moral concerns.^[1] Commons has long held that files that pose such legal or moral concerns should be deleted even if they would otherwise be within the project's scope.^[2]

This guideline applies only to images of identifiable people; it does not apply to generic representations of humans. The identifiable person does not need to be named in the prompt for the guideline to apply.^[3]

Original images

Images of identifiable people created by AI are not allowed on Commons unless at least one of the following criteria are met:

The image in question was published by the person it depicts
The image in question is the subject of non-trivial coverage by reliable sources^[4]

For the purpose of this guideline, "people" refers to any identifiable human, living or dead, that is not conclusively known to be fictitious. In cases where there is scholarly doubt about whether a historical figure actually existed or not (as is the case with many notable individuals in national and religious histories), this guideline treats them as non-fictitious.^[5]

Such images must also be within the project's scope and can't be derivative works of copyrighted images.

Altered images

Generative artificial intelligence is unable to reliably upscale small photographs or colorize black and white photographs without introducing significant errors or fabrications.

Photographs of people that are subsequently altered by AI are not allowed on Commons unless all of the following criteria are met:

The original, unaltered image is in the the altered image's {{Information}} template. This can be fulfilled either by uploading the original image to Commons and including in the "other versions" field, or by linking to the external source hosting it in the "source" field.^[6]
The file is tagged with {{AI upscaled}}.^[7]

Such images must also be within the project's scope and can't be derivative works of copyrighted images.

Notes

↑ Issues that have been raised in discussions on Commons include accuracy (AI images can't guarantee, and often fail to, accurately represent the people they purport to depict), dignity (AI images can be created that depict people performing actions or endorsing ideas that the people they purport to depict did not perform or endorse), and privacy (AI images can be used to create images of people that prefer not to be depicted, either by AI specifically or in general).
↑ In other words, even though Commons:Project scope is a policy and Commons:Photographs of identifiable people#Moral issues is a guideline, a file that violates the latter should be deleted even if it's acceptable by the former.
↑ This guideline would not ordinarily apply to an AI image with the prompt "Gandalf fighting the Balrog" because Gandalf is conclusively known to be fictional. However, if the image depicts Ian McKellen as Gandalf, the guideline would apply, even though Ian McKellen's name was not in the prompt.
↑ Commons is not equipped to judge whether a specific source is reliable or specific coverage is substantial, so this is a low bar. Galleries or BuzzFeed-style listicles of interesting images would not meet this standard, but most other coverage would.
↑ For example, there is significant scholarly doubt whether Moses existed, but this guideline covers them. The guideline is this broad because there are already ample freely licensed, non-AI works depicting such figures, and Commons does not want to become a venue for people debating the historical veracity of these individuals.
↑ This serves three purposes: it demonstrates that the file in question is an alteration rather than a new generation, it allows others to verify that the source file is available under a free license, and it gives anyone using the altered image an easy way to find the original and make an informed decision about the quality of the alteration.
↑ Even though this won't be a perfect fit for all cases, this is currently the only template on Commons that denotes that an image has been altered by generative artificial intelligence.

[1] Issues that have been raised in discussions on Commons include accuracy (AI images can't guarantee, and often fail to, accurately represent the people they purport to depict), dignity (AI images can be created that depict people performing actions or endorsing ideas that the people they purport to depict did not perform or endorse), and privacy (AI images can be used to create images of people that prefer not to be depicted, either by AI specifically or in general).

[2] In other words, even though Commons:Project scope is a policy and Commons:Photographs of identifiable people#Moral issues is a guideline, a file that violates the latter should be deleted even if it's acceptable by the former.

[3] This guideline would not ordinarily apply to an AI image with the prompt "Gandalf fighting the Balrog" because Gandalf is conclusively known to be fictional. However, if the image depicts Ian McKellen as Gandalf, the guideline would apply, even though Ian McKellen's name was not in the prompt.

[4] Commons is not equipped to judge whether a specific source is reliable or specific coverage is substantial, so this is a low bar. Galleries or BuzzFeed-style listicles of interesting images would not meet this standard, but most other coverage would.

[5] For example, there is significant scholarly doubt whether Moses existed, but this guideline covers them. The guideline is this broad because there are already ample freely licensed, non-AI works depicting such figures, and Commons does not want to become a venue for people debating the historical veracity of these individuals.

[6] This serves three purposes: it demonstrates that the file in question is an alteration rather than a new generation, it allows others to verify that the source file is available under a free license, and it gives anyone using the altered image an easy way to find the original and make an informed decision about the quality of the alteration.

[7] Even though this won't be a perfect fit for all cases, this is currently the only template on Commons that denotes that an image has been altered by generative artificial intelligence.

[1]

[2]

[3]

[4]

[5]

[6]

[7]