Talk:Community Wishlist/W253

Semi-automatic for testing

Maybe this should be added to the proposal but at first / if there are complications it may be a good idea to make this possible in a semi-automated way like buttons to Scan this whole category+5 layers of subcategories via image reverse search and Scan this user's uploads for copyvios with each file getting a new datetime field for last tineye/GIRsearch (it could be stored in the structured data or separately with the UID/URL of the file) so it's not scanned twice (for some time at least since rescanning can also be a good idea). Prototyperspective (talk) 09:16, 18 August 2024 (UTC)

Implementation details are over-specific

Hi @Prototyperspective I think your proposal here is too specific. Reverse image search isn't the only way we might do copyvio detection, and if this was to be done by the foundation it probably wouldn't be via a "bot"

I strongly agree that automatic detection of copyvios in uploaded media is something that would be very useful, but that gets lost a bit in the details of this wish. Would you consider editing it to focus more on the problem rather than the solution?

CParle (WMF) (talk) 15:35, 31 July 2025 (UTC)

I don't see why. That there are also other means of detecting copyvios doesn't mean a wish about having that is overspecific. Why would it not be done by a bot? The point is that it's done automatically, so a bot I think would be part of the process...e.g. for adding a hidden category like "Likely copyvios to check" that users can quickly check via some tool.

Again, I don't see why and what would need rescoping and/or editing. Prototyperspective (talk) 10:58, 9 August 2025 (UTC)

@Prototyperspective: reading between the line here, I believe what CParle (WMF) is saying is that if you want WMF resources for copyvio detection, you might get them, but not for this particular proposal. Since he holds pretty much all the power in this situation, it seems futile to argue with him about that. Is there any reason not to submit a different wishlist item simply asking for help with the problem, rather that a particular tool he seems to be rejecting as not in his (or WMF's) purview? - Jmabel (talk) 06:51, 5 December 2025 (UTC)

The whole point of the proposal is automatic copyright detection which can only work by crosschecking via the image i.e. image reverse search. I don't know what an other proposal would be. It may be interesting and useful if CParle elaborated. Either way, so far (1) automatic bot/script copyright violation detection is the whole point here because that would alleviate the huge workload on the volunteers of something that can be done automatically (2) the wishlist isn't just for the U.S. WMF but for the entire Wikimedia movement so for example volunteer developers could also implement this. Nevertheless, if you or CParle have any ideas for another approach, I'd be interested. Note that CParle did not elaborate why he thinks my proposal is too specific. Prototyperspective (talk) 12:24, 5 December 2025 (UTC)