This is an archive of past discussions. Do not edit the contents of this page. If you wish to start a new discussion or revive an old one, please do so on the current talk page.
Movie player
Latest comment: 1 month ago5 comments3 people in discussion
@DoveTheEditor: Hi, and welcome. You want to keep it, then? Your language is confusing. Who snapped the photo? How did you acquire the copyright, since you wrote that it is your own work? You may need to use the Monty Python quote "I'm not dead yet". See also COM:HR. — 🇺🇦Jeff G. ツ please ping or talk to me🇺🇦 14:15, 1 May 2026 (UTC)
I've tagged the enwiki sandbox for speedy deletion there as a blatant hoax. (It claimed, among other things, that she died last year and was awarded the Presidential Medal of Freedom.) Omphalographer (talk) 15:50, 1 May 2026 (UTC)
Looks like a hoax all around, and a pretty transparent one at that. This may be a reason to block the account. - Jmabel ! talk17:03, 1 May 2026 (UTC)
This section is resolved and can be archived. If you disagree, replace this template with your comment. Prototyperspective (talk) 16:58, 4 May 2026 (UTC)
Unmerge files / Undo overwrite
Latest comment: 1 month ago3 comments2 people in discussion
@Nakonana: The first thing to do is a revert (done), then downloading the second version and re-uploading with a new file name, using the diff as licensing evidence. As I'm currently using my tablet, I can't do that. Regards, Grand-Duc (talk) 19:42, 2 May 2026 (UTC)
Thanks for the instructions. I've uploaded the other file under a different name (got the year wrong, it's 2016, not 2015). Nakonana (talk) 22:03, 2 May 2026 (UTC)
This section is resolved and can be archived. If you disagree, replace this template with your comment. Prototyperspective (talk) 17:00, 4 May 2026 (UTC)
Latest comment: 1 month ago1 comment1 person in discussion
There are currently 3,416 files on Commons which are categorized generically as "art". This is a lot - particularly given that the Art category is a disambiguation and should be empty!
Please help us make this number smaller by:
Removing the category from files which already have more appropriate categories.
Diffusing files to more specific categories, e.g. to categories for the artist, or for the subject matter of the work.
Nominating files for deletion which are unlikely to be of educational use. There is a lot of personal artwork in this category which can probably be speedily deleted as F10.
Latest comment: 27 days ago12 comments6 people in discussion
Lately, I have been having trouble with Google Lens and TinEye checking for copyright issues with images. For example, with File:MISS MONDE.jpg "Something went wrong. No image at the URL. Try again with a different URL or image" in Google Lens, and "TinEye could not read that image url. This may be due to an unsupported file format". What could be the cause? I use Firefox 140.10 on Mac. Wouter (talk) 19:04, 4 May 2026 (UTC)
You can also go into Preference > Gadgets, and enable the "Reverse Image Search" tool, which will place automatic search links to common sites in tabs above the image (different skins may have them in other locations). — Huntster (t@c)20:03, 4 May 2026 (UTC)
@Alexis Jazz: It seems that they settled on only using certain sizes for performance reasons, and our gadget doesn't use any of them. Our gadget needs to be fixed. — 🇺🇦Jeff G. ツ please ping or talk to me🇺🇦 20:19, 4 May 2026 (UTC)
This section is resolved and can be archived. If you disagree, replace this template with your comment. Prototyperspective (talk) 14:17, 7 May 2026 (UTC)
Unexpected category related to my own uploads
Latest comment: 28 days ago4 comments3 people in discussion
I just ran across Category:Uploaded by user Jmabel, which has apparently existed for a decade and of which I was never informed. It contains 653 files, which appear to be some arbitrary subset of my probably 3000 to 6000 Flickr uploads (many of which are my own work first uploaded to Flickr, many of which are from Seattle Municipal Archives Flickr stream, and probably 100-200 of which are from other sources). I never asked for this category, I never was informed of this category, and the only users to edit it appear to be a a bot that has not edited in years, a user who has not edited in years, and a blocked user I have some vague memories of interacting with, but not about this. Can anyone tell me what is going on here and whether this is useful to anyone? (I have some other questions, but they'd depend on the answers to those.) - Jmabel ! talk06:35, 5 May 2026 (UTC)
Comment, I found discussions related to this in the bot's request page and a Village Pump thread from 2015. Basically, per the closing comment of the Village Pump thread, this user category was created to "delete uploader information from the source field" and the uploader information is "converted to an uploader category". Thanks. Tvpuppy (talk) 09:28, 5 May 2026 (UTC)
I think the above info solves this. Probably, it would be fine if the user that the user-cat is about empties the cat using cat-a-lot and then has it deleted. If in doubt, one could create a CfD for it. Prototyperspective (talk) 12:31, 5 May 2026 (UTC)
I see. So someone unilaterally decided that for bot-uploaded files that were uploaded at the request of a particular user, the only connection between that user and the file would be a maintenance category that they didn't name clearly (nothing related to the bot-upload aspect) or explain in a hat-note on the category, nor did they explain this to the users in question, nor get their consent to remove the explicit mentions that were previously there. Brilliant.
This section is resolved and can be archived. If you disagree, replace this template with your comment. Prototyperspective (talk) 14:17, 7 May 2026 (UTC)
Featured galleries?
Latest comment: 1 month ago4 comments2 people in discussion
I see that the page Commons:Featured galleries proposed featured galleries nearly 20 years ago. It seems like it just slipped through the cracks and never got made? I for one would love to see this become an actual feature. I have been working on improving a lot of galleries and great galleries are few and far between — I think they ought to be recognised. Plus it would be very helpful to have community consensus on what makes for a top-notch gallery.
I think more focus needs to be placed on galleries, since they are the main namespace, and the most accessible part of Commons for casual users. LetmeEditit (talk) 18:24, 1 May 2026 (UTC)
@LetmeEditit: you may be a mile ahead of me here, but if you wanted to propose guidelines for featured galleries, you might extrapolate from Commons:Galleries, including from some of the examples there of good galleries. Also, if you think there are better examples, that would be helpful. - Jmabel ! talk01:08, 2 May 2026 (UTC)
I'll start putting some guidelines together now; when I'm done I'll put them up on Commons talk:Featured galleries. In fact, there's some old discussion on that page that may be useful.
I think the guidelines should be pretty open-ended, considering a lot can be done with galleries, so I think being too rigid would restrict creativity. I think something similar to the featured picture guidelines might be good — some sort of a vote based off personal preferences. Do you have any ideas of your own? LetmeEditit (talk) 11:48, 2 May 2026 (UTC)
Latest comment: 1 month ago1 comment1 person in discussion
Net sheds at Fishermen's Terminal, Seattle
Category:Net sheds is redirected to Category:Fishing huts, but that seems absurd for something like these industrial-scale net sheds at Fishermen's Terminal on the south side of Salmon Bay in Seattle. I would imagine other major fishing fleets have something similar, though if we have photos of them I have no guess how someone chose to categorize them. - Jmabel ! talk04:45, 3 May 2026 (UTC)
Introducing WISE: Semantic search for Commons (we’d love your feedback)
Latest comment: 26 days ago12 comments10 people in discussion
This is a search tool for images and videos on wikimedia. Currently is searching only on media of the day (approximately 5000 videos). It searches only the visual content of the file (not on metadata, filename, or structured data). It also does face detection and recognition.
Please give it a go. Here's some taste example queries:
for visual queries (select "Visual" on the dropdown menu next to the search box)
"man at a train station"
"एक व्यक्ति रेलवे स्टेशन पर" # this is "man at a train station" in Hindi (this tool handles multiple languages, so try on your language)
"horse in an airplane" # the first result is correct but it's a ogv file
విమానంలో గుర్రం # this is "a horse in aeroplane" in Telugu (again, try it on your own language)
man with a flower
pirate with a pistol
for face search (select "Faces" on the dropdown menu next to the search box. Then click on the "Image" green button, and past the URL for an image)
[warning: some videos are ogv videos and your brownser may not play them, so watch them on Commons, i.e., after clicking the search result, scroll down to the media metadata table, and follow the link to commons]
Key features include:
Semantic search using natural language to find relevant images on Commons from the visual content only (not the structured data or description, only the image itself)
Face search: upload or paste a face image, and the tool will try to identify and locate that person across images and videos, including timestamps where they appear
Audio search: search within audio files to find relevant segments
Multilingual search: supports queries in multiple languages
We are actively improving the tool and would really appreciate any feedback, suggestions, or ideas from the community. Here's some more immediate future plans:
Search on all wikimedia images (instead of only videos / media of the day)
Show similar images when uploading an image to suggest filenames, categories, and other relevant metadata
@Gopavasanth: aren't there privacy issues around doing a face search? I would think that if applied to (for example) crowd photos at political demonstrations, the results could be pretty terrifying. - Jmabel ! talk04:41, 3 May 2026 (UTC)
Face search without identifying person, there is person in photo, there are some persons in photo, there is lot of persons in the photo or generic there is males/females/childrens/adults/eldery people in the photo could be more useful and most likely more robust than trying to identify a specific person. --Zache (talk) 09:46, 3 May 2026 (UTC)
+1 to Jmabel. I would strongly oppose any tool that integrated facial recognition into search capacity. 19h00s (talk) 12:23, 3 May 2026 (UTC)
I mean, if you are concerned about political repression, the police almost certainly have much better tools then this running on publicly available images on the internet. Bawolff (talk) 16:56, 3 May 2026 (UTC)
We can not prevent this and anyone should be aware of this when publishing photos. But we should not host such a tool, that also clearly would violate EU regulation, on the Wikimedia infrastructure. GPSLeo (talk) 13:06, 7 May 2026 (UTC)
This is cool and all, but I'm not sure its very useful unless it covers all the images (or at least a much larger subset). At the same time, it seems unclear there is a path forward to actually doing that. Bawolff (talk) 16:57, 3 May 2026 (UTC)
Great initiative and a very compelling use of embeddings. I was wondering if you've got any ideas as to how to benchmark the retrieval quality when the corpus grows? Awinkler3 (talk) 07:57, 4 May 2026 (UTC)
It's exciting to see projects that make it easier to discover the great contents from Wikimedia projects.
I wanted to introduce a recent initiative that may be relevant for projects like Wise that are based on Wikimedia content. The Wikimedia Attribution Framework has recently been launched, and we are looking for early adopters to learn from their experience.
The Wikimedia Attribution Framework sets guidelines on how to provide sustainable attribution when reusing Wikimedia content. It is in an early beta stage, and we want to learn from those trying to apply the guidelines provided. We'll be adding more details to the project page, but feel free to share any thoughts on the talk page.
I hope this could be a useful resource to make the Wise project even better.
Note also that the Attribution API is also available to facilitate the application of the guidelines provided in the Attribution Framework. For the case of WISE, it can be particularly useful to show license info for images.
Nice, thanks for developing this! I wonder if categories set on items could be used as clues as to the content of the files to improve performance and also whether this tool could be used to basically suggest categories that are missing (eg major content of video but no related set cat found). Prototyperspective (talk) 14:19, 7 May 2026 (UTC)
Issues with FileImporter
Latest comment: 26 days ago2 comments2 people in discussion
If you know of more precise categories for the images (and considering the available subcategories, this is quite likely), please do. --HyperGaruda (talk) 19:14, 8 May 2026 (UTC)
I would suggest to check out other comparable communes (like Rome or Milan?) and copy elements of the category structures there, to reach the level of precision you intend.
Some words of caution: Categories are intended to group similar things together, but in doing so, they may create uninteded divisions in other aspects. This file could be placed in a category for the specific church, but would then also need to be put into "Sculptures in San Polo" (or "...in Venice") and maybe "San Polo in the 2020s". If you create new and super-specialized categories, or chose names that deviate from regular category names, you should be able to explain your ideas in a debate. --Enyavar (talk) 21:39, 8 May 2026 (UTC)
New footage from Cleopatra (1917 film)
Latest comment: 24 days ago1 comment1 person in discussion
A couple years ago someone discovered 40 seconds of film from the famous lost 1917 film, Cleopatra, starring Theda Bara. This film is sometimes cited as the most expensive lost film of all time. Surprisingly, no one has uploaded this clip to Commons. The new footage can be found at https://www.youtube.com/watch?v=QwPZuyF2Th0. However, someone will need to remove the new audio track and the introduction. Unfortunately, the uploader decided to put a watermark on the footage, but I don't think there's anything we can do about that. It doesn't affect its public domain status regardless. Nosferattus (talk) 20:50, 9 May 2026 (UTC)
Template:Incorrect Exif date
Latest comment: 23 days ago5 comments3 people in discussion
i have some photos that have the correct date but wrong time. technically the current phrasing of Template:Invalid Exif date "Date and/or time...are incorrect" is not applicable to my files. is there another template for right date wrong time? or should this template be modified?
I can see in the documentation, the supposedly second way has been struck and commented out. The text of the second way states, "Give the time difference of Exif as a parameter. The template displays the corrected date (localized by {{ISOdate}}) and renders the words "according to Exif metadata (corrected)" in the language specified in the user's preferences. (not available yet)".
As this method is "not available yet", actually only one method is possible, so I have edited the documentation to remove the confusing text. Thanks. Tvpuppy (talk) 23:41, 5 May 2026 (UTC)
that offers a clue. Sarang might have planned to introduce the "time difference" parameter.
i think, calculating the difference and then letting the template add that to exif time, is not a practical solution to correct the error.
i'm gonna change the wording to "Date, time, or both...are incorrect" so the template also works for right date wrong time.
Требую по-русски объяснить, по какой причине были удалены сделанные мною фотографии объектов и архивных документов.
Я, Максим Догадин, русский, гражданин России, нахожусь на территории России и общаюсь на русском языке.
Maximdogadin (talk) 22:33, 9 May 2026 (UTC)
Там разные причины удаления от разных файлов, главные состоят в том, что есть сомнения, что Вы являетесь правообладателем на те фотографии, которые загружали, а некоторые файлы нам просто не подходят. Ymblanter (talk) 06:43, 10 May 2026 (UTC)
I demand an explanation in Russian as to why the photographs of objects and archival documents that I took were deleted. I, Maksim Dogadin, am Russian, a citizen of Russia, I am on the territory of Russia and communicate in Russian.
For what it's worth, various files were deleted for various reasons. It would be good if a Russian-speaker would take this on, but it's probably not going to be one single answer. - Jmabel ! talk02:44, 10 May 2026 (UTC)
Как я вижу, всё что угодно было загружено как собственные работы, тогда как в действительности это по большей части файлы, взятые откуда=то из Интернета. Andrei Romanenko (talk) 10:09, 11 May 2026 (UTC)
Здравствуйте. Под собственной работой сюда можно грузить только то что сами сфотографировали. Чтобы не было сомнений в вашем авторстве, загрузите фото в оригинальном размере и с метаданными. Но файл может быть производной работой от другого произведения, например памятник, мурал или документ. Тогда нужно изучить правила COM:DW и COM:FOP. Даже если файл находится в общественном достоянии, нельзя писать "собственная работа" как вы сделали тут, пусть даже вы сканировали сами. По возможности, автора нужно указать. Дату сканирования нужно заменить на дату создания работы, как например, здесь. Чужие работы допускается грузить только если есть явное согласие автора на одну из разрешенных лицензий. Юрий Д.К.11:05, 11 May 2026 (UTC)
Courtesy deletion requests and reasonable expectations from nominator
Latest comment: 21 days ago32 comments10 people in discussion
Looking at the recent courtesy DRs i noticed that in almost every case the nominator insists that us hosting the photo is violating their legal rights and privacy yet they never seem to provide any reasoning for that accusation or even what caused them to make the accusation at all.
I know we are just deleting them out of courtesy but i still feel like we should expect something more substantial than "This photo is illegal because i say so!" reasoning that seems to plague these kind of DRs
Where do we draw the line when it comes to expectations from the nominator? Is there just simply not one? Trade (talk) 22:35, 8 May 2026 (UTC)
I know "GDPR" have gotten it's share of criticism but the whole "no one is allowed to show me face on the internet" doesn't seem entirely accurate to the actual legislation Trade (talk) 23:06, 8 May 2026 (UTC)
I think it's hard to directly "draw a line" there or make any strict rules about it; it always depends case-by-case. I personally think that we can be quite lenient towards granting such requests, especially if the images in question are unused and/or low quality and/or easily replaceable . A common misconception in "Keep" arguments in such discussions seems to be "we won't delete it, because CC licences are non-revocable"; while that is true, it dosen't mean we can't delete the image - deleting the image won't affect it's licence, we're simply no longer distributing it under the licence. Other people can still continue using the image under the terms of that licence, even if it's no longer hosted here. ~TheImaCow (talk) 22:48, 8 May 2026 (UTC)
But surely we should be allowed to expect a reason at all, no? And what if the claims made by the nominator doesn't stand up to scrutiny? Like the constant insistence that they never gave any permission for us to use their photo when they are the author and we can clearly see that they did indeed uploaded the photo under a free license Trade (talk) 23:03, 8 May 2026 (UTC)
Examples please. Privacy boundaries within public places or events, like a parade or carnival, might be within a grey area. George Ho (talk) 02:26, 9 May 2026 (UTC)
@George Ho: Going through my contributions for the past year:
In Commons:Deletion requests/File:Gio Porta.jpg, a TA asked for removal of a selfie for privacy and as uploader and as oos in Italian. Italy presumed to be country of complaint. I !voted to delete. Ongoing.
In Commons:Deletion requests/File:Image of yuvadeepthi Smym Mammood.png, the uploader requested deletion for privacy as a part of being renamed away from a spammy username per enwiki. I fixed the subpage after an errant tag, but I didn't !vote before Herbythyme deleted. Malaysia presumed to be country of complaint.
In Commons:Deletion requests/File:Lisa Ullyna Tarigan dan Antoni Albalat di Jakarta.jpg, a TA claiming to be a notable poet's ex-wife asked for deletion of a two-shot including her and the poet for privacy reasons. We have a crop of just the poet, which relies on the two-shot for provenance. I asked about keeping that crop. Indonesia or Catalonia (Spain) presumed to be country of complaint. Ongoing.
In User talk:Boyprostido, the user never managed to create the subpage, despite tagging their user talk page in an L2 header citing "Personal information, incorrect identity, and privacy concerns". See my discussions with that user on that page, where they basically admitted to sockpuppetry over an extended period of time, revealing usernames that had promoted their music and music empire while explaining their motives. Nigeria presumed to be country of complaint. See en:special:diff/1315969765 for background.
In Commons:Deletion requests/File:三上功太.jpg, the pictured uploader requested deletion for personal privacy reasons. I !voted weak delete. The Squirrel Conspiracy deleted for F10, despite half a dozen live edits in mainspace on jawiki. Japan presumed to be country of complaint.
In Commons:Deletion requests/File:YQgNKUvs.jpg, the "model changed her mind", presumably for her privacy. I !voted to delete and IronGargoyle deleted. Nearly a year later, the uploader was indeffed as a VOA. No country presumed to be country of complaint.
In Commons talk:Abuse filter/Archive 2025/10#Report by Morvran teardrop, the user wished "to remove all information relating to it from Wikipedia for copyright reasons and to allow my projects to evolve" per Google Translate. I explained that they had irrevocably licensed (among other things) and they did not respond. France presumed to be country of complaint.
I dunno which ones were taken at a public setting. Nonetheless, re- reading the OP, this is about self-portraits, right? George Ho (talk) 06:00, 9 May 2026 (UTC)
I think we do and we should always delete selfies on request. But of course the request needs to come from the uploader. All other requests need to go through a VRT process, as we can not check the identity of the nominator. GPSLeo (talk) 06:41, 9 May 2026 (UTC)
This is meant to be about courtesy nominations in general and what is and isn't acceptable behavior from the nominator Trade (talk) 17:34, 9 May 2026 (UTC)
In cases where someone claims their legal or privacy rights are being violated when they technically are not, I don't think we should get too hung up on the legalese. Subjects nominating photos of themselves for deletion are more often than not unaware of Commons policies and relevant laws. What matters more is that the subject does not want the photo up, and we should balance that request against the needs and scope of our projects. Was the subject aware they're being photographed? Is the photo bad or the subject being portrayed in an embarrassing situation (e.g. mid-yawn)? Is the subject notable on Wikipedia or Wikidata? Do we already have other images that can serve the same purpose? Does the photographer/uploader agree with the removal? Could broader consequences come from keeping the photo up (e.g. the photographer getting in trouble with a venue or event)? Those are the kinds of questions I think we should be thinking about with these requests, regardless of the exact justification a nominator gives. ~Kevin Payravi (talk) 18:10, 9 May 2026 (UTC)
But should they even be required to give any justification in the first place? It's hard for me to take a stance on courtesy DRs when we are essentially barely given any context or anything to work withTrade (talk) 18:45, 9 May 2026 (UTC)
@Trade: It is very helpful if they mention something about privacy at the outset, rather than starting with empty "reason=" or "reason=reason". I patrol incomplete deletion requests. — 🇺🇦Jeff G. ツ please ping or talk to me🇺🇦 18:51, 9 May 2026 (UTC)
[Tha]t is helpful but should providing a reason be a requirement? [A]nd the ones that simply consist of "Remove or "I don't want this here anymore"? Trade (talk) 18:54, 9 May 2026 (UTC)
@Trade: reason is required by {{Delete}}. If there is no valid reason, I look for a reason to delete anyway; that's how I find some copyvios (and tag them accordingly). With no imputable reason, I revert as incomplete. I warn as and when necessary. — 🇺🇦Jeff G. ツ please ping or talk to me🇺🇦 19:02, 9 May 2026 (UTC)
So you are telling me if the person who wants the file deleted haven't provided any reason we should add the DR to Category:Incomplete deletion requests? Trade (talk) 00:29, 12 May 2026 (UTC)
@Trade: If they don't provide a reason parameter (or provide an empty one) in the delete tag, the tagged page will be automatically categorized there; the subpage will not. — 🇺🇦Jeff G. ツ please ping or talk to me🇺🇦 15:39, 12 May 2026 (UTC)
"In cases where someone claims their legal or privacy rights are being violated when they technically are not, I don't think we should get too hung up on the legalese" There is a very high chance that these claims ends up running afoul of COM:NLT though Trade (talk) 18:56, 9 May 2026 (UTC)
That's not what either COM:VRT or meta:VRTS say. Both have words to the affect: "[VRT] handle queries, complaints, and comments from the public by email".
The meta page discusses licensing, then continues, "There are many other queues in use within our VRTS implementation. They range in purpose but share one core purpose: to assist in facilitating communication amongst Wikimedia users, readers, customers and anybody else who has something to say!" with the Commons page saying "The main use of VRTS in relation to Commons is to verify and archive licensing permissions.". "Main use" clearly signifies "not the only use". Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits16:22, 12 May 2026 (UTC)
I remember that some (self-claimed) owners of houses demanded deletion of pictures of their buildings in Germany. The images showed cultural heritage monuments and were mostly (probably even all) taken from public ground, without revealing sensitive information like names. They were kept usually, as the "Recht am Bild der eigenen Sache" does not exist in Germany. Many people with no are almost no legal background expect that they can decide who is allowed to take pictures of their property, but this is not true in that absolute case --PantheraLeo1359531 😺 (talk) 14:25, 9 May 2026 (UTC)
I remember a case where a person from a political discussion panel requested a deletion. This was valid in my mind, as the focus was on this person, and politics can be controversal. In Germany may other legal elements apply than in the US --PantheraLeo1359531 😺 (talk) 15:31, 9 May 2026 (UTC)
So let's consider this one I closed a few hours ago. I might even have supported a polite request for a courtesy deletion (not a very good photo, but in use, we'd have needed to find a substitute), but the user was making legal threats. As I said in my closing remarks, "[I] certainly see no reason to grant a courtesy deletion in the face of what looks to me like a poorly grounded legal threat." (That was probably the least polite thing I said in a paragraph of closing remarks.) - Jmabel ! talk02:30, 10 May 2026 (UTC)
As for Commons:Courtesy deletions i am not really a huge fan of "Please give more detail than "I don't want this here anymore"." It makes it sound like something optional rather than a requirement users are being expected to follow--Trade (talk) 00:49, 12 May 2026 (UTC)
Arbitrary sorting with numbers for categories
Latest comment: 21 days ago6 comments2 people in discussion
@Jeff G.: well, I thought it was obvious that I am asking people what they think, if they agree with what I am saying or not :-) maybe it does not necessarily need to be in COM:RFC, otherwise the talk pages may lose part of their usefulness. It was a matter of categories, so I discussed in the talk of the pages about categories. But I see what you say, in the future I will double check where to start a discussion, thanks. --Superchilum(talk to me!)19:48, 12 May 2026 (UTC)
Latest comment: 21 days ago4 comments2 people in discussion
I would like to have my userpage deleted so that my MetaWiki page shows up in all the projects. At the same time, I would be happy if it could be kept (with all the history) as a subpage (like User:Oudeís/Archive). Unlike Wikipedia, Commons does not allow me to move the page with Special:MovePage. Is there anything I can do about it? Are subpages allowed at Commons in the first place? Oudeís·talk18:45, 12 May 2026 (UTC)
Possibly, Commons:Paying public domain (especially if the landmarks are cultural heritage works whose architects or sculptors have died many many years ago)
Should Wikimedia Commons have a separate page discussing privacy rules and/or "house rules" of property owners, especially those related to indoors of establishments that are already in the public domain? Or should someone expand the relevant section at "COM:Non-copyright restrictions" instead?
Note that I started the discussion here, as it is not about prohibitions imposed by painters, sculptors, architects, or the copyright laws. It is about restrictions imposed by property owners or property managers of various places or establishments, many of which that are in the public domain because their designers have died for more than 50/70/100 years.
Do note that COM:CRSM might no longer be true. COM:CRSM claims Wikimedia Commons has no obligation to honor privacy rules: "Photographs taken by yourself in a museum or the interior of a building/monument are deemed acceptable here, provided they do not show copyrighted works. If the museum's house rules forbid photography, a breach of that rule is an issue between the photographer and the museum, but does not affect the copyright status of an image. If the museum's house rules were a valid contract, it would bind only the parties of the contract: the photographer and the museum. Wikimedia Commons and all other third parties are not subject to such a contract." JWilz12345(Talk|Contributions)13:04, 10 May 2026 (UTC)
We should not mix house rules and or breaking of some kind of (informal) NDA with privacy problems. They often overlap but are something totally different. This example is not about personal information of a real person. We can discuss, when we should delete files based on house rules or non disclosure agreements, but we should not discuss this under the label "privacy". GPSLeo (talk) 15:13, 10 May 2026 (UTC)
Also, I looked at the Brazilian church deletion request thread, and what I would say in regard to it is: If you are given a tour of a building under the condition that you won't post photos online and agree to that condition, don't post photos online. Had they been COM:INUSE, the issue would have been more complicated, but since they were not, it was easy to treat the deletion request as a courtesy to the church. I doubt that we should create a broader policy than that. -- Ikan Kekek (talk) 20:02, 10 May 2026 (UTC)
I think this particular case was precisely a privacy issue, even if the privacy was that of an institution (the church) rather than an individual. This is very different than (for example) a museum that wishes to prevent photographs of its non-copyrighted artworks so that it can better exploit them economically. - Jmabel ! talk19:58, 10 May 2026 (UTC)
This is information that they want to be treated confidentially for religious or security reasons. For me the term privacy only applies to information about identifiable individuals, but never institutions. GPSLeo (talk) 20:24, 10 May 2026 (UTC)
Here, I would apply the notion of privacy even though this does not involve a natural person. The purpose of privacy protections is often connected to security: publishing certain information can make someone more vulnerable to malicious actors. Highly visible people that already live with limited privacy usually also have the resources and infrastructure needed to manage those risks.
Thinking from this perspective, a world-famous church such as Notre-Dame has extensive visibility and security infrastructure to match. On the other hand, a historically valuable church located in a remote area and with limited resources may become significantly more vulnerable if sensitive interior details are widely shared online. In that sense, even if "privacy" is traditionally associated with individuals, I think it is reasonable to recognize that certain institutions may also have legitimate concerns that Commons should take into consideration. ━ AlbertoleoncioWho, me?11:44, 13 May 2026 (UTC)
Could you take that to Commons:Village pump/Proposals please (once Bawolff is done with adapting user:Putnik's existing gadget)? Default implementation of blurring was at no part of the property proposal discussion and i dont wish to give the appearance that Commons have already agreed to it Trade (talk) 02:03, 12 May 2026 (UTC)
@Trade You added the same value for real photos of people and for illustrations. If both of these files are marked in the same way, we can not filter as we wanted to. GPSLeo (talk) 16:44, 12 May 2026 (UTC)
I never said anything about the items being limited to photos or illustrations during the property proposal? Not sure where you got that from Trade (talk) 21:01, 12 May 2026 (UTC)
The idea of this tool was to allow people to filter specific content. If two content types with different filter demands have the same tag, we can not separate these two types anymore. GPSLeo (talk) 06:00, 13 May 2026 (UTC)
I suggest the rule that if an image has it classification changed 3 times in a year it foreever loses the ability to have a classification attached to it ;) —TheDJ (talk • contribs) 22:02, 12 May 2026 (UTC)
Latest comment: 20 days ago3 comments3 people in discussion
Hello from the Product Safety and Integrity team! After a successful trial of hCaptcha for bot detection on English, French, Japanese, and four other Wikipedias, we will be rolling out hCaptcha to all projects over the course of the next few weeks.
Specifically, this will include using hCaptcha for account creation, and for edits by newer users. This replaces the use of the traditional "type in the word" CAPTCHA that is currently required of many users for these flows. hCaptcha is set up to only challenge likely-suspicious activity, so very few humans will be interrupted at all. In our trial, we estimate that only about ~2% of human users were actually challenged.
For background - as part of our focus on securing the wikis and detecting bad-faith activity, we are building stronger protections against bots carrying out activities that are generally intended for humans, such as creating accounts and editing. As the web has evolved and computers have gotten smarter, our old CAPTCHA has become both challenging for humans and easy for computers.
We recognized this and began with a trial of hCaptcha, third-party bot detection service, to replace our old CAPTCHA. Since then, eight Wikipedias, including English, French, and Japanese, have been using hCaptcha to protect account creation and certain kinds of edits. hCaptcha is a company that specializes in bot detection, with experience protecting very large online websites while prioritizing user privacy in its design. We have implemented technical safeguards to reduce the sensitivity of the information we send to hCaptcha.
It also gives us "likely bot" signals for account creation and covered edits, regardless of whether the session was challenged. Those signals inform what we show CheckUsers and stewards, who use them to find and remove bad-faith activity that was likely done by bots and which may not have been found any other way.
@SGrabarczuk (WMF) We have one filter that uses a captcha to limit spam that was typically done manually in the web browser. I would assume that in such cases the new captcha does not work anymore to limit this type of spam. GPSLeo (talk) 06:05, 13 May 2026 (UTC)
@GPSLeo The AbuseFilter showcaptcha consequence will continue to function as intended, with an "always challenge" mode being used when it is triggered. KHarlan (WMF) (talk) 14:16, 13 May 2026 (UTC)
Location around Copenhagen 2003
Latest comment: 19 days ago3 comments2 people in discussion
Thanks. With hindsight it was logical that when I took pictures of metroline 1 (File:Kopenhagen metro 2003 3.jpg), I ended up at the train station of Ørestad. I am stil hesitating between Litra ET and Littera X31 multiple units.Smiley.toerist (talk) 10:03, 14 May 2026 (UTC)
Ladies lounges
Latest comment: 19 days ago3 comments2 people in discussion
Ladies' lounge of the Liberty Theatre, Seattle, circa 1914
Certainly similar, but distinct. These have largely fallen into abeyance, although some still exist, mainly in older buildings (mainly theaters, sometimes hotels or ballrooms). Women-only spaces, usually doubling as a passage from the public circulating space to the women's rest room in the sense now used. These were still pretty common in New York City when I was a child (1950s, early 1960s). There was rarely any equivalent for men, and if there was it was on a much smaller scale. I would think it deserves a category. - Jmabel ! talk19:15, 14 May 2026 (UTC)
Bot request
Latest comment: 19 days ago4 comments2 people in discussion
Latest comment: 18 days ago10 comments4 people in discussion
The particular case is Commons:Deletion requests/Files uploaded by Linna222 but some opinions on the issue in general are welcome. What we have here are a bunch of photos and documents, seemingly from 1900 or so. Copyright isn't the real issue here, but all these files were uploaded as "own work", which is impossible as the oldest person ever was 122 years old. If these files were high-resolution scans with EXIF scanner information, one could argue these came from a family or municipality archive or something. Instead, we get limited resolution (~1000 pixels) images, many with visible compression artifacts and/or scaling and the source for all of them is the same: they are screenshots. Screenshots of what? The uploader explained nothing. The provenance trail is utterly non-existent. In the uploader's defense, Wikimedia set them up for failure. But that's a story for another day. I generally AGF and if the uploader had said "I got it from this Facebook page/family archive/newspaper article" I would have probably shrugged. What gets my goat is that there is literally no statement from the uploader whatsoever and meanwhile User:Richard Arthur Norton (1958- ) inserted {{Scan}} ("Scan from the original work") as the source in all but two of the files. Projects like Wikipedia shouldn't accept this. What about Commons? - Alexis Jazzping plz06:09, 13 May 2026 (UTC)
for this particular case.
typical stuff related to china: sometimes getting reposted over and over on chinese websites and the source is either never acknowledged, or quickly lost due to linkrot or china's closed intranet.
you can probably sieve out something even within commons though:
i just tested what it looked like in the visualeditor.
i think its instruction is simple and clear. the user is to blame in this case. they just clicked "This is my own work" even though it's not. also instructed to follow "If you do not own the copyright on this file, or you wish to release it under a different license, consider using the Commons Upload Wizard." but the user didnt even though they obviously "do not own the copyright on this file". RoyZuo (talk) 13:09, 15 May 2026 (UTC)
Every newbie does the same thing. You wrote: "the user is to blame in this case", I disagree, the instructions are not clear when you scan or take a photo of an existing work. The form should have clear instructions to credit the original work and the new derivative work. --RAN (talk) 17:13, 15 May 2026 (UTC)
COM:EVID says that provenience is typically requiring a stated source and author to be specified. That is not prescribing the only and absolute way to provide evidence of PD, as far as I interpret the policy.
Take File:Miensk. Менск (1751-1800).jpg. In that case, the "source" indicated is basically fully worthless; authorship is unknown; and even image searches won't find the original file because it was screenshotted and maximaly cropped. But from the drawing style, I have no doubt that the indicated age range is correct and that the pd-old-100-license was applied correctly. My interpretation of COM:EVID is that old age sufficiently demonstrates PD. My alarm bells on COM:PCP begin ringing once I read dates ranging from the 1920s and later; in those cases we just cannot allow sloppy documentation.
Please note that unclear provenience IS a big problem which we should not simply dismiss. I hate dealing with material where the uploaders don't know authors and claim ridiculous ages ("between 1700s and 1800s" is ~300 year range once you begin taking into account illiteracy). Even if we know that an image is PD-old and within our project scope; crappy documentation by the uploader can render files essentially worthless for educative purposes. We need to educate uploaders on how to best document the sources and authors; and when they begin uploading a LOT we really need to sensitize them on the issue. All that said, not everyone of our contributors is a born archivist: We should accept low-accuracy (i.e. sloppy) documentation with provable PD files. When doubt arises, of course, we have to act according to PCP. --Enyavar (talk) 10:45, 13 May 2026 (UTC)
The uploader has not responded but most likely had access to the John Lawrence Thurston papers at Yale while researching the topic for the Wikipedia article. The archive has three photographs of Thurston, and she uploaded all three. Yale says: "much of the material in this collection is likely in the public domain" which is the same conclusion we have come to at the deletion nomination page. See: https://archives.yale.edu/repositories/12/resources/2971 for more information. --RAN (talk) 04:45, 14 May 2026 (UTC)
Can I publish my own university transcripts on Wikimedia Commons
Latest comment: 18 days ago7 comments5 people in discussion
This includes heavy hard-baked redactions to prevent exposing personal information as well as exclude any elements that might meet the threshold for creativity.
Below the threshold of originality because it's probably somewhat like an index with no lengthy prose and hardly any full sentences. Nakonana (talk) 14:46, 15 May 2026 (UTC)
I'd stop short of saying transcripts are in scope broadly. There's the copyright issue being discussed, but I can't imagine us needing random people's (no offense!) transcripts. If it's an excellent quality scan that illustrates the concept of "transcript" then sure, I suppose, but I wouldn't generalize to say "yes we love transcripts!" :) — Rhododendritestalk | 18:20, 15 May 2026 (UTC)
Hi @Jmbael: : Intended in-scope use would be to have generic examples of various modern university transcripts (note that the ones currently in https://commons.wikimedia.org/wiki/Category:Transcript_certificates are all almost 100 years old or more). There are currently none at all on Wikimedia Commons, for the last decades. I have highly redacted the transcripts, such that no copyrighted material (logos, signatures, etc... are included) and almost 0 full sentences are included, hence it would fall out of scope of copyright protection. Also, I would like to include it on my user page. Vincent Mia Edie Verheyen (talk) 21:32, 15 May 2026 (UTC)
I think the right thing to do is probably to upload it, with the understanding that (like almost any file) it could imaginably be deleted. Certainly you won't be in any trouble for doing so. - Jmabel ! talk01:45, 16 May 2026 (UTC)
Free PDF maker that maps the text
Latest comment: 17 days ago4 comments3 people in discussion
When Adobe Acrobat runs OCR it stores the text in the PDF and maps the text to the image, so that when you right click and hold, you are grabbing the text from the image. Is there free software the performs the same function? RAN (talk) 17:03, 15 May 2026 (UTC)
Thanks! I will give it a try. Have you tried making djvu files yet? Equally perplexing. --RAN (talk) 15:51, 16 May 2026 (UTC)
I used OCRmyPDF when cleaning up/retouching this book. The results are somewhat mixed - most of the text is accurate, but it recognizes and generates nonsense text in B&W illustrations that isn't there, and I haven't found a way to edit inaccuracies (such as the description below the first color illustration) manually. ReneeWrites (talk) 17:33, 16 May 2026 (UTC)
Filippo Ferri copyright state
Latest comment: 11 days ago7 comments4 people in discussion
Hi, what's the correct copyright template for this picture? On the site, is written CC BY-NC-ND-2.5 IT, but I can't find the template. --Giorgx12 (talk) 13:08, 22 May 2026 (UTC)
@Giorgx12 There is no template matching that license on Commons, because that license is not compatible with Commons (that license does not allow derivatives or commercial use). 19h00s (talk) 14:02, 22 May 2026 (UTC)
Latest comment: 16 days ago7 comments5 people in discussion
We started in February 2025 to categorise the 125,000 files in Category:All media needing categories as of 2021. Three quarters of the work have been done, but now the question arises: How to categorise the last quarter? I think, we need experienced volunteers, please, unless we prefer to wait for AI or useful bots:
Do you have an idea, how to tackle this task, or could you even help us please, by manually categorizing some files?
NearEMPTiness (talk) 04:40, 2 May 2026 (UTC)
Only the suggestion that people try starting at pseudo-random places in the sequence. I was just able to knock off 5 of them pretty easily in about 10 minutes, down in the second page of the S's. I think there's (surprisingly) still a lot of reasonably low-hanging fruit no one has rally looked at. - Jmabel ! talk19:04, 2 May 2026 (UTC)
Jmabel is right: There are still quite some self-explaining images that can quickly be categorised. I can confirm from my experience with "letter D" files.
On the other hand, obviously there are also images where you've got an idea (or just a whiff) of what they probably show ... and then, it takes half an hour of work to pin them down.
Identifying "nests" of "same target category" images helps both speeding up and keeping the categorisation consistent. With files that have "serial" file names, this is rather obvious. Less obvious: When an uncategorised file is used in a Wikipedia article (such as w:fr:Michel Biot, in my case), there may be more uncategorised files in the same article. -- Martinus KE (talk) 09:27, 5 May 2026 (UTC)
Now down to 11,900 images, not bad progress. And every so often you find some gems in there.
This 1897 poster had been sitting uncategorized.
A little bit of work with GIMP and I got this out of it.
A large number of these images are copyright violations, low quality stuff, spam, personal photos, out of scope material, etc., so don't hesitate do nominate such material for deletion. Unfortunaly there are many people simply mass-moving huge numbers of images into various top-level categories such as Category:Diagrams, Category:Logos, Category:Unidentified people, Category:Men, Category:Drawings, Category:Documents, various country-categories, Category:Buildings, Category:Screenshots, etc, causing those to be absolutely flooded with random material, which someone has to sort out later. This dosen't really solve anything, it simply moves the problem somewhere else, and it is especially problematic if a lot of the material has very questionable licencing. ~TheImaCow (talk) 16:49, 17 May 2026 (UTC)
@Draceane: Because categorization is tricky, and relatively new users (especially people who are here only because they need to upload an image in service of a sister project) need to have a relatively easy workflow. Also, some very valuable bot uploads from GLAMs don't have information that maps well onto our categories, and they often can't do much better than "bot upload from such-and-such archive, needs category check". For example, I'm most of the way through categorizing and otherwise curating Category:Images from the Prosch Albums, very useful stuff for Seattle history, and BMacZero did a good job uploading them for me but, honestly, other than Category:Images from the Prosch Albums to check, most of the categories in the initial upload were useless at best. Lots of Category:Business districts, lots of Category:Seattle (too generic, in any case, and often wrong), etc. No problem, I think, in this case because they were uploaded precisely with the understanding that I would work on them once the were uploaded. - Jmabel ! talk21:49, 17 May 2026 (UTC)
GSoC 2026: Lossless JPG Transformations Project
Latest comment: 15 days ago15 comments6 people in discussion
Hi Everyone!
I am Rishan, an Open source contributor. I have been contributing to the Wikimedia Commons App for the last couple of months and I am excited to share that I have been accepted as a Google Summer of Code(GSoC) contributor at Wikimedia Foundation for this summer. I want to take this opportunity to share about the upcoming improvements in the image editing capabilities of the app.
The Wikimedia Commons App is an Android application developed by contributors across the wikimedia movement that enables users to browse, upload, and contribute media directly to the Wikimedia Commons repository using their mobile devices.
This Project under GSoC 2026 aims to implementing the following:
1) Lossless crop feature.
2) Lossless blur feature.
3) Auto-detect and apply blur to people's faces and car number plates.
1) Lossless crop feature.
The Existing Crop feature in the Commons app when applied results in loss of information. Due to improper handling of METADATA.
The Goal: Implement a "smart" crop that preserves all original METADATA while maintaining the highest possible image resolution. This ensures that when a user crops a photo to focus on a specific subject and the resulting image is remained as its original resolution.
2) Lossless blur feature
Sometimes while uploading the photo, the contributor would want to blur out some parts of the image to preserve privacy.
The Goal: A lossless blur feature, which would let contributors blur out the regions which they don't want to be seen when uploaded, to make it lossless we apply blur to only regions required by the user and make sure we don't recompress other blocks of the image.
3) Auto-detect and apply blur to people's faces and car number plates.
When contributors capture images in public places, they often capture faces of bystanders or car number plates that won’t have privacy clearance for a global platform like Wikimedia Commons.
The Goal: Using computer vision library OpenCV, the app can automatically scan the image to identify the coordinates of human faces and car number plates these coordinates can be used to blur human faces or car number plates out of the image.
Outcomes:
No loss in information after editing the picture.
No need to depend on third-party apps instead transformation is done in the right way.
Protecting people's privacy becomes easy.
This summer, I’m honored to be working under the mentorship of Nicolas Raoul and Ritika Pahwa. Excited to learn from the best!
That is cool, but i'm not sure blurring people's faces is exactly something we want to encourage except in exceptional circumstances. Bawolff (talk) 15:45, 9 May 2026 (UTC)
I think it's a good option to have. Some countries seem to be very strict about privacy questions like this. Nakonana (talk) 15:55, 9 May 2026 (UTC)
Hi @Jmabel, No it doesn't trigger as long as the user doesn't opt for it, As for how it gets to know which faces to blur is that when user triggers this option, It utilizes the OpenCV computer vision library, which uses specialized algorithms that scan the entire image and return a list of coordinates for every face identified. Those coordinates are then passed on to the blur feature to apply blur to only those region of the image. This ensures that only the identified faces are blurred while the rest of the image remains untouched Mechi4895 (talk) 19:19, 12 May 2026 (UTC)
Yes, I understood that. That is not what I asked. Let me go back to the example I linked above: File:Snoqualmie Moondance drummers 02.jpg. This shows three well-known performers plus some other people who happened to be with them in a drum circle. Do I understand correctly that the user would have no way to choose which people's faces would be blurred, and that it would blur even Lansing Scott, who is in sharp focus at right, and that there is no way for a user to control that? - Jmabel ! talk21:02, 12 May 2026 (UTC)
The user will be able to manually blur what they want, if they don't like the result of the automatic detection. Syced (talk) 14:06, 13 May 2026 (UTC)
Hi @Nakonana, You are very right about countries having strict privacy and we have only kept it as an option, which is not triggered by default. Mechi4895 (talk) 19:11, 12 May 2026 (UTC)
Hi @Bawolff, No we are not encouraging and making every image uploaded go through this process.
One note, blurring can be partially reversible in cases when a simple blurring algo is used. So might wanna keep that in mind. —TheDJ (talk • contribs) 22:10, 12 May 2026 (UTC)
You are right, and to avoid taking any risk with this, we are thinking of making the whole blurred area single-color (average color of the area pre-blurring). What do you think? Syced (talk) 14:04, 13 May 2026 (UTC)
I concurr and see what you mean! Don't know how I didn't notice that at the time. Have now brightened and reduced the shadowing to lessen the grey tube-like appearance. Hopefully it looks better, otherwise if consensus is to revert to original no qualms with that. :) ⟦ Lιᐰndrєι ⟧ ✧ talk15:35, 23 May 2026 (UTC)
I think it's good now, oppose revert to original. I consider proper background removal minor under COM:OW. More importantly, it's a clear improvement to all the many, many pages where it's used. (Why is the v2 staying cached, despite a refresh? odd! ) Uploader User:Longhair? -RememberOrwell (talk) 16:52, 23 May 2026 (UTC)
Latest comment: 9 days ago2 comments2 people in discussion
This section is resolved and can be archived. If you disagree, replace this template with your comment. ReneeWrites (talk) 10:42, 24 May 2026 (UTC)
Hello. I am trying to seek advice as to how it would be possible to move these two categories (1 and 2) to their previous titles. They were moved back in September 2023 without any discussion and the Wikipedia entries for the English edition as well as other editions show that the previous names were indeed the common names. I would appreciate it if a more experienced user could clarify the process on the Commons for me as it appears to be different from how it's done on Wikipedia. Thank you. Keivan.fTalk07:18, 24 May 2026 (UTC)
Maybe could use quarry to look for pdf files where the first page has different dimensions than the other pages, and the conversion program in the metadata is google books. Personally, what i wish for was a way to tell mediawiki to use a different page than page 1 as the default page. Bawolff (talk) 15:09, 18 May 2026 (UTC)
@Yann, (from a sample size of a few, anyways) the cover-sheeted scans all have one of two thumbnails for page 1 of the PDF. It should be possible to scrape the thumbnails for matches
This Quarry query picks up some files with a lot of noise. This got a bunch more of them, but doesn't perfectly filter out stuff.
Overall, I think scraping the 1st page thumbnails and checking them against the known cover-sheets would work. JayCubby (talk) 14:46, 19 May 2026 (UTC)
I meant a quary more like [2] which i believe should get most of what you want without noise. (Unfortunately its slow, I guess WMF removed the index on mime type). Bawolff (talk) 00:06, 20 May 2026 (UTC)
commonsarchive dot org still desired or required
Latest comment: 14 days ago5 comments5 people in discussion
Hi, there once was a project at Wikimedia Cloud services using a custom domain commonsarchive(dot)org. The project goal was to archive raw material e.g. from digital cameras. After the domain expired I grabbed (unfortunately missing to talking with Steinsplitter beforehand who attempted the same) it and set its A records back to the Wikimedia Cloud servers. Is the Commons Archive project still alive? If so, would it like me to update domain records or get the domain name transferred? Thanks in advance. Rillke(q?)08:39, 19 May 2026 (UTC)
FileImporter still fails to automatically tag local images on enWiki with w:en:Template:Now Commons tag. It has been almost 14 days since I first reported it there, yet it hasn't been fixed.
It appears FileImporter can now automatically add such tags, starting with some of my file transfers of Patrick Roque's drone shots of some parts of Metro Manila. However, I remain vigilant due to the tendency of some mediawiki tools to "break" or "fail" in recent months. JWilz12345(Talk|Contributions)11:36, 20 May 2026 (UTC)
Earliest file created by Commons user
Latest comment: 13 days ago7 comments3 people in discussion
Just wondering, which Commons user uploaded their original works that were created the earliest? Which users are the top 10? And if we extend to people who might not have operated a Commons account by themselves but donated files expressly to Commons?
Probably Commons users who have been photographing in the 1950s, or even the 1940s / 1930s? 15:56, 20 May 2026 (UTC)
RoyZuo, I misread, but this is also interesting. revision 1 (14:37, 7 September 2004) just boringly says "First post!". Less than half an hour later the main page was vandalized in the second revision. Uploading apparently didn't work yet as revision 6 states "Please mail your images to tstarlingwikimedia. o r g. He knows how to use Python and will upload them for you" -- Kate ({{@}} and thin spaces added against harvesters) The first revision that shows anything file-related is revision 25, the creation of User:Node ue which transcludes Image:Quail1.PNG uploaded 19:52, 7 September 2004. I'm not sure if that's also the "earliest file created by Commons user" in the sense of the first ever uploaded file. The oldest original work from a Commons user you'll never find. I regularly find works from 18xx claimed as own work. You'd have to determine what the oldest yet legitimate claim is. A 100-year old person who uploaded in 2004 would have been 15 in 1919, if one of their parents was a photographer they might have taught their kid how to take a photo. - Alexis Jazzping plz17:16, 20 May 2026 (UTC)
He might be the person who posted the oldest user-generated content online, that photo of him (if it was a selfie, not his dad's photo). To beat him someone has to have made something before 1937 (photos, music by music prodigies, paintings by art prodigies...), live long enough to use the internet, and post their works somehow publicly. I think most probable people to do that are musicians, if someone wrote something before 1937, and published that by themselves or instructed people to do that online. Ennio Morricone "wrote his first compositions when he was six years old" (1934/35).
Some of us were uploading images to en:w before the existence of Commons. (Unfortunately, often deliberately as small size before the "thumbnail" feature was introduced.) -- Infrogmation of New Orleans (talk) 18:51, 20 May 2026 (UTC)
May 2026 Wikimedia Café meetups regarding the Wikimedia Foundation Annual Plan
Latest comment: 12 days ago1 comment1 person in discussion
Yeah, that may be true for the cywiki version of the template, but its equivalent on Commons requires a US template. Nakonana (talk) 16:18, 25 May 2026 (UTC)
I think temporary adding a Commons license template to the cywiki file description might do the trick. It will probably break the cywiki display of the license but might be enough to trick the tool into thinking that the license is Commons-compatible. After exporting the file you could then revert the license change on cywiki. Nakonana (talk) 16:25, 25 May 2026 (UTC)
Latest comment: 11 days ago4 comments3 people in discussion
Upon the recent death of Brazilian footballer Geovani Silva, I have came to notice that his main portrait available on Commons looks poorly AI upscaled:
Original image on Placar : After some investigation, it looks to me like the original picture is this one, in a poor quality but depicting a man with a different face than the upscaled one.
Since I'm not a regular Wikimedia Commons contributor, I don't really know the usual stance on this matter (I don't even know if this discussion should take place at the Village pump or somewhere else), but from my understanding the use of AI upscaling in Commons is controversial (Commons:Village_pump/Archive/2026/02#More_explicit_policy_against_upscaling_needed).
What actions should be taken in that case? Should we leave the AI upscaled portrait as it is, or replace it by the original? Or maybe leave both versions co-exist?
Latest comment: 3 days ago7 comments3 people in discussion
Hello! I have translated the Igbo version of the Main Page at Main Page/ig. Please i want an admin to link this page to the correct global Wikidata item for the Commons Main Page, and update the main language template so that Igbo will officially appear in the language switcher list at the bottom of the page Wmbata (talk) 10:32, 30 May 2026 (UTC)
I'm not at all sure how the Main Page is internationalized, does someone understand the mechanism? - Jmabel ! talk14:06, 30 May 2026 (UTC)
Latest comment: 8 days ago4 comments3 people in discussion
The file en:File:Logo of Restore Britain.png is marked as non-free content on the English-language Wikipedia. Surely this is free content, as it consists only of text and a map outline in which there cannot be any copyright, as the outline of the map is determined only by coastline (not a human work, therefore no copyright), and political boundaries, specifically the boundaries of counties of Ireland, made well over 100 years ago and thus out of copyright even if one had existed in the first place. The Anome (talk) 12:26, 25 May 2026 (UTC)
Most UK party logos may only be used under fair use, which is not acceptable for Commons. Labour is just as unfree, for example.
Regarding THIS logo: A specific generalization (but not the landform itself!) may be copyrighted. As maps go, this one is on the very low end of intellectual property if any, but PCP shoud be applied. I would think that Restore Britain might have registered a trademark on this general color/map/text combination/arrangement. Unless there is a valid carveout that states why Copyright may not apply. --Enyavar (talk) 20:02, 25 May 2026 (UTC)
The fact that it's a party logo is irrelevant, as is the status of other party logos. Either it's copyrightable, or it isn't.
Latest comment: 7 days ago32 comments10 people in discussion
This proposal has been re-drafted endlessly. It's difficult. Before I actually put this up on COM:VPP I'm sharing it here in case it contains serious flaws.
As it would affect a longstanding policy I intend to request a project wide banner to make the user base aware of the proposal.
The current draft:
Misleading and misinforming content
Works where a reasonable person is likely to make incorrect assumptions about its provenance when only looking at the work (not the file page or file name) [Note 1] are not acceptable.
Visual works that were modified using AI tools (e.g. restoration) can not be acceptable unless:
It is ensured by an experienced user that the AI did not add artificial details outside of areas that had to be filled like creases or watermarks.
The unmodified original is uploaded alongside the modified file.
Details of any AI tool use are detailed on the file page.[Note 2]
Human-modified visual works that misrepresent the original are not acceptable.
Works where the educational value is in the misinformation and optical illusions are exempt.[Note 3] Files can be cured by superimposing information that declares the provenance[Note 4], provided that it is legible at the default thumbnail size.
Footnotes
↑Like mistaking an AI-generated image for an actual photograph or historical artwork. This does not affect generic drawings and paintings where a reasonable person would not attribute them to a particular author or studio.
↑For example in the upload comment, using {{Retouched picture}} or (if available for the tool/operation in question) categories.
↑Will Smith eating spaghetti is an example which was covered by multiple news outlets. Another example would be a well documented art forgery. This does not extend to AI-generated/modified works made by Commons users to demonstrate flaws of AI, those should be cured with superimposed provenance.
↑The language of choice should be appropriate for the intended audience of the file. For AI-generated works the initialism "AI" is always sufficient. For files uploaded before $proposal_acceptance_date there is a grace period until $proposal_acceptance_date+1_year to allow uploaders and projects to make their files compliant or upload them locally.
localized imperfections I'm not sure "localized" is the key. Maybe "specific identified"? For example, I could imagine a good AI based tool to deal with despeckling, half-toning, inadvertant moire, etc., certainly one at least as good as "conventional" tools for this in Photoshop or GIMP. - Jmabel ! talk21:56, 17 May 2026 (UTC)
definitely better. Wondering about "which are detailed on the file page". Seems to me that things like removal of a watermark aren't usually noted on the file page when done by any other means, unless you count the upload comment for an overwrite as a "mention". I don't see why doing it with AI would require different handling. The two images are still there to compare, and if it had other side effects then it is a bad job and should be reverted. - Jmabel ! talk01:05, 18 May 2026 (UTC)
Jmabel, when a human does it there are typically no severe side effects. With AI, unless it's tailored to that specific purpose, it's guaranteed to mess it up. If it's not detailed on the file page we wouldn't be aware that we should check. For anyone who's unaware: when you say "fix red eyes in the image", non-tailored AI doesn't desaturate a few pixels. Oversimplified: it converts the image to "George Clooney wearing a blue-brown vest with round buttons at sundown facing the camera dreamy look with red eyes and baseball cap yada yada yada", removes the "with red eyes" part as per your instructions and uses that prompt to render a new image. I once asked Bing to remove the black fat bike from File:Police bicycle in 2025 in the Netherlands side 1.jpg. And upon my request, Bing did pinky promise it wouldn't change any pixels that didn't belong to the fat bike. Then it mangled the police bike, turned all the signs into gibberish, converted the w:Albert Heijn logo to "a1" (if you squint the AH logo does look a little like "a1"..), made modern art out of the bike rack that passed through the shop window, moved the walls, changed the lighting, and so on. Bing has not been a good Bing and I gave it a good spanking. If everybody complied with all the rules, you'd be right. In practice it's more messy, and given AI's predisposition to messing up, redundancy is not a luxury. Some users will forego the review by an experienced user, some will skip the superimposed information, some will omit their prompt from the file page. Compare with gun safety: some rules will be broken, but you generally must break more than one to have an accident. For example, when you holster a handgun it's pointing at your feet, but you should never allow the muzzle to point at something you're not willing to destroy. If you follow the rule of not putting your finger on the trigger until you've made the decision to shoot, your feet will be fine. - Alexis Jazzping plz09:09, 18 May 2026 (UTC)
unless it's tailored to that specific purpose: exactly. Nothing in the current wording addresses that. - Jmabel ! talk23:53, 18 May 2026 (UTC)
@Jmabel: I've changed the second point per you and GPSLeo, but I don't see a way to differentiate between non-tailored AI and AI tailored to a specific purpose. Even when it is tailored, quality will vary. This would require maintaining a whitelist of acceptable AI tools. - Alexis Jazzping plz16:44, 19 May 2026 (UTC)
This is not a review comment, but I would create a new section at Commons talk:Project scope to discuss this, or even a new RFC page, rather than VP and VPP. I think it's better to use a dedicated discussion space when you anticipate a big discussion than a discussion space for miscelleneous topics. (We can use VP and/or VPP to advertise it, though, but if you are planning a sitenotice, that alone might be enough.) whym (talk) 23:04, 17 May 2026 (UTC)
@Whym: , there was one singular RfC in 2025 and it has 2 whole comments. RfCs aren't used much on Commons, it seems. But with a sitenotice and VP/VPP notices an RfC may indeed be better. A sitenotice would be appropriate in my opinion as this proposal would affect basically all Wikimedia projects. - Alexis Jazzping plz23:27, 17 May 2026 (UTC)
A few comments:
What is the point about "human-modified visual works that misrepresent the original" meant to accomplish?
In footnote 1, can we add something to the effect of "an actual photograph or a piece of historical artwork"? I've seen the latter a number of times.
Footnotes 2 and 3 should be rolled into the text of the bullet point and made more clear.
In footnote 4, the "does not apply..." language seems unnecessary. Some projects legitimately use user-generated AI images to illustrate AI errors, like File:AI generated woman with extra legs.jpg; I don't see any compelling need to prohibit that.
Thank you for your suggestions! What is the point about "human-modified visual works that misrepresent the original" meant to accomplish?
Essentially codifies cases like Jan Arkesteijn's block into policy. He claimed until the very end he had no idea why he was being blocked.
Footnotes 1, 2 and 3: done.
I don't see any compelling need to prohibit that.
It wouldn't be. The draft has three bullet points, only the first applies to images generated out of thin air. No reasonable person would be confused about the provenance of File:AI generated woman with extra legs.jpg as her anatomy is really impossible. This also isn't meant to apply to AI-drawings, watercolor paintings and the like. For example this drawing (if it was in use) isn't meant to be targeted by this as a reasonable person wouldn't mistake it for a Vermeer, work from Studio Ghibli or a photograph. The language of footnote 4 (now footnote 2) would only apply to demonstrations of more subtle errors, like a hand with six fingers which is actually plausible. Such images would still not be prohibited, they would only have to be cured with superimposed information to avoid invalidating its COM:INUSE defense. - Alexis Jazzping plz10:04, 18 May 2026 (UTC)
I would change the second point. The current formulation does not cover the core problem well. I would simply say: "Images modified using AI tools, if it is not ensured that they did not add artificial details." This makes clear that edits like de-noising with suitable tools are not a problem. GPSLeo (talk) 06:17, 19 May 2026 (UTC)
@GPSLeo: when you remove watermarks/creases/etc you essentially always introduce artificial details, regardless of method. I've reworded it, is this better? I'm saying "visual works" instead of "images" because AI does video too. - Alexis Jazzping plz16:44, 19 May 2026 (UTC)
I don't fully understand all implications of this proposal and of where the proposed insertion is going to be. Does this mean that provably misleading/false/forged files could be subject to deletion? If so, I could support this, I have long combatted fraudulent productions - not so much fake photos, but faked graphics. --Enyavar (talk) 21:06, 19 May 2026 (UTC)
Example of superimposed provenance@Enyavar: Good point! As the intention is for these new rules to be able to override COM:INUSE I thought they should be added under that section, but that would have been a mistake. The draft has been adjusted accordingly. Some implications:
Examples at w:AI upscaling would require superimposed provenance.
Anything photorealistic that COM:AIP disapproves of would be automatically out of scope, even if INUSE, unless provenance is superimposed. (may still be deleted for other reasons, like being an unused non-notable artwork)
Nothing happens to w:Madame Tussauds. A reasonable person may make incorrect assumptions about the provenance, but that's also where the educational value lies.
Nothing happens to File:Jesus Trump.jpg as it was covered by various news outlets, and even it hadn't been, nobody would be confused about its provenance. Everybody knows Trump is not a doctor.
Your unmodified original is uploaded alongside requirement here differs from COM:AIIP which currently permits a milder or by linking to the external source option. I'm not sure on the figure, but a lot of images in Category:Upscaling (such as upscaled YouTube stills) do not include an unmodified copy on Commons. Belbury (talk) 15:28, 22 May 2026 (UTC)
Belbury, considering the prevalence of linkrot that's a flaw of AIP that shouldn't be repeated in this proposal. a lot of images in Category:Upscaling (such as upscaled YouTube stills) do not include an unmodified copy on Commons. That's an absolute shame, but strictly for this proposal in its current form it wouldn't matter. AI-upscaling, unless cured with superimposed provenance, is not acceptable. No matter if the original is uploaded or not. Uploading the original only helps (in the sense of not needing superimposed provenance) if an AI-tool was used for something other than upscaling (for example watermark removal), checked by an experienced user, the AI-tool use is detailed on the file page, and the unmodified original is uploaded. Note that non-AI upscaling isn't affected in any way. - Alexis Jazzping plz16:20, 22 May 2026 (UTC)
Oppose This is unrelated to COM:INUSE. If an image is in use in another project it's on scope. If the image is wrong on some account it should be marked as such in Commons or it even can be removed from categories, but it can't be deleted as out of scope without first it stopping being in use on the other project. This seems just another proposal aimed at making Wikipedias store locally the images they use instead of using Commons, as it was before Commons was created.--Pere prlpz (talk) 12:48, 21 May 2026 (UTC)
Strong oppose Commons should not become an arbiter of what is "misinformation", "misleading" or sufficiently "factual". That is the role of the individual Wikimedia projects through their own content policies, sourcing standards and editorial processes. Commons should only concerns itself with copyright, licensing and basic project scope, and files that are legitimately in use on Wikimedia projects are, by longstanding principle, and should continue to be, in scope. --Jonatan Svensson Glad (talk) 16:51, 22 May 2026 (UTC)
Josve05a, don't you think that ship had sailed when Ratify Commons:AI images of identifiable people as a guideline passed and COM:AIP became a guideline? Because several admins are now overruling COM:INUSE policy by referring to that guideline. In part, this proposal draft is damage control. One of its goals is to make COM:AIP partially redundant. The proposal draft above doesn't strictly ban any content as it allows anything to be cured with superimposed provenance. This way, projects can upload anything, no matter how abhorrent we think it is, as long as it's INUSE and (if it matches one of the bullet points above) has superimposed provenance. This relieves Commons from the burden of hosting the kind of misinformation that many users here object to. If another projects really wants to it could even use w:Template:CSS crop to get rid of the provenance in their local thumbnails! - Alexis Jazzping plz17:31, 22 May 2026 (UTC)
@Alexis Jazz: COM:AIP is about legal and moral rights of subjects depicted. If admins were misusing it to delete images that don't affect legal or moral rights of subjects depicted, we would have a bigger problem.
@Alexis Jazz: No, I don't think that ship has sailed, and even if parts of it have, we should not accelerate it. The fundamental principle of Commons has long been that if a file is legitimately in use on another Wikimedia project, it is in scope. We are a media repository, not an editorial board or truth ministry for the movement. Individual projects (especially enwp) have their own policies on reliability, due weight, misinformation, and image use. That division of responsibility exists for good reason. COM:AIP was (controversially) passed specifically around legal/moral rights of identifiable living people in AI-generated images, i.e. a narrow carve-out. Using it (or this new proposal) as a general tool to delete or sideline files that are merely "misleading about provenance" or "potentially confusing" is a massive expansion. Many historical images, political cartoons, propaganda, satire, and restored works are in some sense "misleading about provenance" or altered. We have never required "superimposed provenance" labels before.
This proposal would put Commons admins in the position of constantly judging "does this look too much like a Studio Ghibli style?" or "is this upscaling good enough?", effectively encourage projects to host files locally again, and lead to endless deletion arguments about what constitutes sufficient "superimposed information".
If another project is using it, it should stay (barring legal requirements not to do so). If the content is misleading, the project using it should be responsible for context, captions, disclaimers, or removal. Not Commons playing upstream censor. If there are specific bad cases that are abusing COM:INUSE, let's address those narrowly rather than creating a broad new exception that weakens one of our core policies. (Please also note difference between a guideline approval and a policy change)--Jonatan Svensson Glad (talk) 18:05, 22 May 2026 (UTC)
Many historical images, political cartoons, propaganda, satire, and restored works are in some sense "misleading about provenance" or altered. Those would generally be covered by the "educational value is in the misinformation" exemption. If you have specific examples that you worry wouldn't be covered by that, I could see if the draft needs to be refined further. Pere prlpz, Are admins using COM:AIP to delete images that don't affect legal or moral rights of the depicted subjects? Josve05a, passed specifically around legal/moral rights of identifiable living people in AI-generated images, i.e. a narrow carve-out.
That is a problem. Then, we shouldn't modify COM:INUSE but COM:AIP to say that it doesn't overrule COM:INUSE.
Or maybe we should modify COM:INUSE to state that it overrules all other policies except those based on legal or moral requirements (like copyright). Pere prlpz (talk) 18:41, 22 May 2026 (UTC)
@Alexis Jazz: You’re overwhelming the discussion with examples in what looks like bad faith. Yes, true legal requirements (copyright, privacy, personality rights of living people, etc.) override COM:INUSE. That’s not in dispute.
The problem is admins stretching COM:AIP — and now this proposal — to delete files that don’t actually break any law or moral rights, just because they dislike the “AI” label, “misleading” provenance, or “lack of education value”. Your pile of examples (Cleopatra, Napoleon, Tutankhamun, etc.) mostly fall into this category: pure editorial overreach on long-dead subjects.
Commons is a media repository, not the movement’s upstream “truth ministry” or provenance police. Individual projects are responsible for context, captions, and reliability. Forcing “superimposed provenance” on everything shifts us into content censorship.
This proposal guts a core policy (COM:INUSE), guarantees endless subjective fights (“Is this too photorealistic?” “Is the label big enough?”), and will drive projects back to local hosting. The “educational value is in the misinformation” carve-out is hopelessly vague.
If COM:AIP is being misused beyond actual legal requirements, fix that — don’t ram through an even broader exception for upscaling, historical reconstructions, satire and more.
Josve05a, you're calling it abuse (and I'm not gonna say you're wrong), but there doesn't seem to be anything that could realistically be done about it. A proposal to treat AIP as a guideline would be nonsense because that's what it is. Dragging admins to AN/U rarely ends with a lovely cup of tea. You voice concerns about this draft for a proposal while we're staring down the barrel of Special:PermanentLink/1220689243#Proposal: Promote Commons:Upscaling to guideline. Subjective fights aren't even the biggest concern, it'll flat out ban various perfectly acceptable practices, and this has been made clear, and users want it to become policy anyway because nobody has written anything better. Individual projects are responsible for context, captions, and reliability. Forcing “superimposed provenance” on everything shifts us into content censorship. Individual projects are not equipped to deal with this, especially the smaller ones. Any many, including enwiki, actually count on Commons to handle these kinds of things. They just assume that what we host is truthful. For text they require reliable sources, but they'll accept a photo taken by Joe Schmoe and documents he scanned and uploaded from his local library. If they have to question and review every single file, they might as well go back to hosting it all themselves. Some Wikipedians -especially on enwiki- are vehemently against transcluding information from Wikidata, specifically because according to them Wikidata can't get its vandalism and sourcing issues under control. Projects don't want to be belittled, you are right about that. But they don't want us to mindlessly host anything either. Maybe we should see what projects think about it all. Forcing “superimposed provenance” on everything shifts us into content censorship. I disagree. In this context, superimposed simply means in-image. That could also be achieved by adding a bar. Which a project technically could hide locally with CSS crop if they hated us. If you run an ad for a financial or medical product you generally have to include some sort of disclaimer. On food packaging you may be required to list the ingredients. To me, that's not really a restriction on freedom of speech. It's not prohibiting you from saying something. - Alexis Jazzping plz22:24, 26 May 2026 (UTC)
@Alexis Jazz: Have you actually asked the wikis in question (like mgwiki or enwiki) what they want from us, or are you just making shit up? Drawings and AI images are made by us and not meant to be taken as real photos, but editors on projects like enwp can differentiate that themselves — we shouldn’t impose upstream ”superimposed provenance” censorship because smaller wikis might be lazy. I’ll abide by community consensus (of course), so I apologize if I come of as too strong, but this the only time I can express my deep dissatisfaction or highlight my interpretation of our policies and role. --Jonatan Svensson Glad (talk) 23:52, 26 May 2026 (UTC)
Also, again. There’s a difference between a guideline and a policy. Policy is always “the highest authority” unless a guideline can shed light on a legal requirement. --Jonatan Svensson Glad (talk) 23:54, 26 May 2026 (UTC)
Josve05a, Have you actually asked the wikis in question (like mgwiki or enwiki) what they want from us, or are you just making shit up? I remember this from a discussion that I probably had on enwiki. I don't know if I could find a link, it must have been years ago. If it would turn out I misremembered I would apologize, but I'm pretty sure. so I apologize if I come of as too strong, but this the only time I can express my deep dissatisfaction or highlight my interpretation of our policies and role. You accuse me of "making shit up". You're burning bridges. I was hoping you'd be willing to collaborate with me to write a better policy proposal, because the community is desperate for some kind of policy. But you disrespect me like this, for no real reason, I'd have expected better from you. Also, again. There’s a difference between a guideline and a policy. Policy is always “the highest authority” Yeah it is!! But if other admins are just going to ignore that and do what they want, guidelines are a higher authority than policy. - Alexis Jazzping plz03:05, 27 May 2026 (UTC)
Josve05a, here's a very clear example of why we need this: Special:ListFiles/Thelezifor. This user uploaded dozens of AI-generated images and many are used on mgwiki, so eradicating all uses will be a considerable amount of work. Nobody on mgwiki looked at this and went "yeah that's totally fine and super educational!", they just didn't know. Which isn't too surprising on a wiki with 112 active users and two admins. When provenance is superimposed, at least those 112 active users have a better chance to spot the rubbish. If provenance isn't superimposed, CommonsDelinker will be doing them a favor. If we strictly follow our current policy we have to either host this crap or remove all uses from mgwiki (and potentially go to edit war with the uploader). - Alexis Jazzping plz14:02, 24 May 2026 (UTC)
Commons is not there to editorialize mgwiki (nor enwiki or any other wiki, for that matter). If you think some user in mgwiki is abusing the system, go to admins there or to stewards, but don't delete images that are in use.
Oppose mostly per Jonatan Svensson Glad. Commons' primary purpose has always been to serve other Wikimedia projects, and we should be careful about restricting files that are COM:INUSE and legally unproblematic. @Alexis Jazz: I understand your point about COM:AIIP, but keeping the slippery slope fallacy in mind, I believe the line can be drawn there. The proposed language, especially at the policy level, is way too broad. I might support this if the language was softened a bit and proposed as a guideline instead, but we kind of already have a proposal for that. Phillipedison1891 (talk) 20:07, 26 May 2026 (UTC)
LLM use in discussion
Latest comment: 7 days ago28 comments15 people in discussion
Support. The only time that someone should be posting a comment generated by some LLM or AI or ML tool is if that person is trying to communicate by translating. In that case, you should probably still post your actual material that you wrote alongside whatever an online translation tool posts to be safe. ―Justin (koavf)❤T☮C☺M☯22:46, 18 May 2026 (UTC)
@Koavf: that presumes you are using the LLM-based tool for at least the bulk of your translation. I'll often use Google Translate in much subtler ways than that: e.g. to see if it can come up with a more apt word than I first chose for something where I'm less than confident, or to verify that my writing in a language where I am decent but short of fluent makes sense by translating back into English, etc. I could give other examples but the short of it is that often it would take two or three times as long to describe my process than it did to do the writing. - Jmabel ! talk02:18, 19 May 2026 (UTC)
@Jmabel: Enwiki's policy says ...that are obviously generated (not merely refined) by a large language model or similar AI technology... Both of your comments here cover situations that would be clearly covered by "merely refined". Pi.1415926535 (talk) 03:29, 19 May 2026 (UTC)
Support. I've seen a lot of LLM-generated comments in deletion discussions (typically by the uploader, opposing deletion), and not once have I seen one that coherently made a relevant argument. I have frequently seen them misstate (or outright fabricate) claims about project policy or copyright law. As Justin mentioned above, users who are not comfortable commenting in English should reply in their native language rather than using LLMs to construct messages for them. Omphalographer (talk) 02:30, 19 May 2026 (UTC)
@Omphalographer: so should I stop attempting to help people on the help desk, etc., in their native language, and just write my replies in English or Spanish, the only languages in which I'm really comfortable? (N.B., I read several other languages well, including at the level of reading numerous books, but I don't express myself fluently, and more often than not need the assistance of translation tools for some word or phrasing.) I think that would be ridiculous. I've seen no signs I've gotten this wrong in any significant way in literally hundreds of times I have done this. - Jmabel ! talk02:37, 19 May 2026 (UTC)
Perhaps I could clarify - users should not feel obligated to write comments in English, particularly if they must rely on a LLM to write a comment on their behalf. I would much rather read a user's own words - in any language, at any level of fluency - than some words an AI put into their mouth. Omphalographer (talk) 06:17, 19 May 2026 (UTC)
@Omphalographer: absolutely, we are in agreement there. I'd much rather read good Spanish/German/Portuguese/French/Romanian (& a few others) than bad English, and for the rest I might as well run them through the translator myself. - Jmabel ! talk20:51, 19 May 2026 (UTC)
Oh, I see. Well... too busy or preoccupied with Wikipedia and real-life stuff right now. I don't mind you going right ahead with page creation. George Ho (talk) 08:36, 19 May 2026 (UTC)
Support LLM use should be banned in discussion. Google Translate is not concerned, but the original language should be included. Yann (talk) 08:38, 19 May 2026 (UTC)
@RoyZuo: so now it cost you 20s, some water and electricity for ChatGPT, a bunch of investor money to fund the data center, insane DRAM prices for everyone, and then someone else has the misfortune of checking ChatGPT's work with a spreadsheet. This doesn't sound like a time saver. And ChatGPT got a failing grade because it can't do addition properly. - Alexis Jazzping plz15:53, 19 May 2026 (UTC)
Summarizing the results of a discussion - especially a long and contentious one - is exactly the sort of task which I would never trust a language model to perform. They consistently perform poorly at this type of task; they frequently misattribute comments, focus on irrelevant asides, and reach "conclusions" which are irrelevant or based on (potentially loud) minority opinions. Omphalographer (talk) 18:34, 19 May 2026 (UTC)
en:WP:LLMTALK leaves AI-refined comments alone, as pointed out above. Aren't the comments found in the section you linked too? Grandmaster Huon said this: "this was written with the help of AI, as it helps me articulate my statement". Or are you saying that they relied on LLM more than they said they did? I'm sure some don't like AI-refined, AI-assisted comments, too, and people can call it out on a case-by-case basis, but it seems that there is no consensus on enwiki to automatically invalidate them.[4]whym (talk) 11:26, 20 May 2026 (UTC)
Oppose We should let people express themselves however they like, even if this includes using LLM. Then we judge on the basis of (and solely on) what they chose to post. Why wouldn't we? Why would we constrain how someone develops their post? Have we ever done such a thing before? Are we to police asking other editors for advice? For copying old posts about similar issues? Automatic translation tools? Why would we justify this new restriction on not only the thought they express, but how they got to that point? They make a post, they become responsible for what that post expresses, we react to that post.
Secondly, the ability to detect use of LLM is itself flawed. On a project where we already have a problem of calls for immediate deletion of 'AI' content, citing policies that don't even exist, it's a problem if we are to begin downgrading our opinion of anything, either content or talk: posts, simply because someone has decided that they're 'AI' and thus of implicitly less worth. Andy Dingley (talk) 21:22, 19 May 2026 (UTC)
The proposal seeks to re-use the existing language from the en-WP guideline page. We're not talking about any suspected AI-use, but "obvious AI argument generation". If this addition to the guidelines gets misused to delete unwelcome voices in a debate, there are obvious remedies, like appealing to administrators. --Enyavar (talk) 22:36, 19 May 2026 (UTC)
The problem is that purely AI generated comments tend to be unnecessary long and verbose and to top it off often miss the point (their contents are irrelevant and don't address the issue at hand). Nakonana (talk) 05:58, 20 May 2026 (UTC)
Is that any more of a problem than human-written comments that are unnecessarily long and verbose and miss the point? I think that is covered by the user being responsible for their comments no matter how they are generated. - Jmabel ! talk13:48, 20 May 2026 (UTC)
In the Wikipedia projects I'd say it is a problem because AI generates walls of text and the longer the thread the more reluctant are people to participate in a discussion. So, it kind of has a "chilling effect". But as for Commons, I've rarely encountered AI generated comments, probably that's because on Commons we don't have to seek consensus over article content etc. all the time. There are just not as many discussions here as elsewhere. Nakonana (talk) 11:41, 21 May 2026 (UTC)
Support LLMs cannot communicate, period. The thread that Apocheir linked is a perfect example. The LLM hallucinated easily-falsifiable accusations, and there's no indication that the other editor was even bothering to read my responses before pasting them into the LLM. Pi.1415926535 (talk) 21:40, 19 May 2026 (UTC)
Support - here, en-WP has the right idea. Post that obviously were generated by AI are typically taking a lot of space without any true argument to them. It should be possible/allowed to mark them as what they are, make these posts collapsible, or in very extreme cases strike them entirely - just like how we would strike other non-constructive contributions on talk pages, like for example fully repetitive walls of texts, or insults. However, there need to be safeguards so that we are not suppressing free speech: This is an international project where writers of all languages are allowed to participate, regularly in auto-translations. If users are flagged/reprimanded falsely for LLM misuse, there should be reasonable ways of recourse, like appealing on the admin noticeboard. And if users are systematically misusing such a new guideline themselves to suppress critics, that should also have consequences. --Enyavar (talk) 22:36, 19 May 2026 (UTC)
If an LLM generates an objectively 'bad' post, e.g. one that is repetitive or hallucinatory, then we should critique it on that basis, i.e. that the results were repetitive or hallucinatory, rather than taking a subjective stance that LLMs and their use are themselves inherently inferior. Andy Dingley (talk) 14:41, 20 May 2026 (UTC)
Support as someone who is neurodivergent and sometimes uses generative AI to get an idea of how to effectively communicate something that's in my head. But I don't copy and paste the response; I use it as a starting point to write my own text in my own words.
All Wikimedia projects are built more or less on consensus and good faith. LLMs are very effective at quickly generating authentic human-sounding communication with little to no effort. It's a perfect tool for trolls, agenda-driven individuals, and other bad actors to compromise our community. If someone is clearly and repetitively posting LLM-generated content in discussions, they absolutely should be blocked as disruptive.
Latest comment: 7 days ago4 comments4 people in discussion
Can anyone elaborate, why are short-links of Google Maps or Google Streetview (links beginning with "goo.gl") prevented from posting on Commons? You can bypass it by using long links from the adress row of your browser, example.
A spam blacklist entry makes no sense if you can bypass it, but more important, what exactly is the point of blocking Google Maps or Google Streetview links from Commons? Please remove this nonsensical block a.s.a.p. Thanks. --A.Savin20:09, 26 May 2026 (UTC)
I think that there's kind of a global rule that URL shorteners (and goo.gl is one of those) are generally forbidden. Of course, Goo.gl itself is mostly benign, but any shortener is liable to hide the original page destination. This is what is to be avoided, thus that policy of "No URL shorteners" is understandable (even if it makes citing Google Maps references bothersome, Google's system itself only shows shortened addresses when you're using the function of sharing a location). Regards, Grand-Duc (talk) 20:25, 26 May 2026 (UTC)
It's only in the global SBL which reads \bgoo\.gl\b(?!/maps\b).*. Some projects like frwiki added goo.gl to their whitelist. --Achim55 (talk) 20:39, 26 May 2026 (UTC)
.svg/20px-Pictogram_voting_comment_(orange).svg.png)
Comment I deleted all files as abuse of COM:WEBHOST, and sent 2 warnings. Yann (talk) 17:23, 1 May 2026 (UTC)
This section is resolved and can be archived. If you disagree, replace this template with your comment. Prototyperspective (talk) 16:58, 4 May 2026 (UTC) 
Redirect {{Incorrect Exif date}} per above, functionally I don't see any substantial difference between the 2 templates. Thanks. Tvpuppy (talk) 11:15, 5 May 2026 (UTC)
.jpg)
This 1897 poster had been sitting uncategorized..jpg/120px-Jossot_-_1897_fertilizer_poster_(perspective_corrected).jpg)
A little bit of work with GIMP and I got this out of it.
.jpg)
Done I restored the two categories, thanks for bringing this to our attention. --ReneeWrites (talk) 10:41, 24 May 2026 (UTC)
wikimedia. o r g. He knows how to use Python and will upload them for you" -- Kate
.jpg){kind=link}
.jpg){kind=link}
.jpg){kind=link}
_interior_with_artwork_blurred.jpg){kind=link}
.pdf/page1-960px-The_Philippic_orations_(IA_philippicoratio02cicegoog).pdf.jpg?utm_source=commons.wikimedia.org&utm_campaign=index&utm_content=thumbnail){kind=link}
.pdf/page1-960px-Portraits_d%27oiseaux%2C_animaux%2C_serpens%2C_herbes%2C_arbres%2C_hommes_et_femmes_d_..._(IA_portraitsdoisea00belogoog).pdf.jpg){kind=link}
.jpg){kind=link}
.jpg){kind=link}
_logo.svg){kind=link}
Oppose This is unrelated to COM:INUSE. If an image is in use in another project it's on scope. If the image is wrong on some account it should be marked as such in Commons or it even can be removed from categories, but it can't be deleted as out of scope without first it stopping being in use on the other project. This seems just another proposal aimed at making Wikipedias store locally the images they use instead of using Commons, as it was before Commons was created.--Pere prlpz (talk) 12:48, 21 May 2026 (UTC)
Strong oppose Commons should not become an arbiter of what is "misinformation", "misleading" or sufficiently "factual". That is the role of the individual Wikimedia projects through their own content policies, sourcing standards and editorial processes. Commons should only concerns itself with copyright, licensing and basic project scope, and files that are legitimately in use on Wikimedia projects are, by longstanding principle, and should continue to be, in scope. --Jonatan Svensson Glad (talk) 16:51, 22 May 2026 (UTC)
![[3]](https://commons.wikimedia.org/w/index.php?title=Commons:Undeletion_requests/Current_requests&oldid=1206588521#File:Unconscience_learning.png){kind=link}
Support. The only time that someone should be posting a comment generated by some LLM or AI or ML tool is if that person is trying to communicate by translating. In that case, you should probably still post your actual material that you wrote alongside whatever an online translation tool posts to be safe. ―Justin (koavf)❤T☮C☺M☯ 22:46, 18 May 2026 (UTC)
