what is going on with tumblr's community labelling?? (12/04/25)

_{read on for further update, written 15/04/25}

i haven't made a blog post like this before but i need to complain and find out if anyone else knows what's going on. it seems that, in the past day or few days maybe, Tumblr's gone crazy with community labelling. some of this makes a bit of sense, for example posts referencing suicide & suicidal intent seem to have been labelled, but on the other hand so have completely innocuous posts. case in point, the pinned post of petslifeworld. now, i'm in no way familiar with this blog, but from what i can tell they simply post cute animal videos. the one in question, as linked above, is a video of a ferret biting (in no way particularly dangerously) a person's hand, before leading the person to where her (the ferret's) children are. when the person moves their hand away, the ferret bites it again to lead them back to the box where the baby ferrets are. its very cute. and yet, here's how it apppears on mobile.

for some reason, on mobile, this harmless and cute video is labelled as "potentially mature content". seriously, what has tumblr flagged here? does it think that the person being bitten is being severely harmed? or are the small squeaks the baby ferrets make too violent?? it's completely unclear, but here's where it gets even more perplexing. below is a screenshot of the same post on the desktop browser.

there's no community label. how does that make sense? my only theories are that either, they've somehow only implemented community labelling on the app to make it more appealling to advertisers, or it could somehow? have something to do with me not having updated the app for the new ui yet. i don't know nearly enough about computers to understand this inconsistency, but it seems really quite concerning. we all know tumblr is notorious for arbitrarily banning trans women for simply existing, so if the mobile app is more representitive of what they're doing with community labelling it makes me seriously concerned for a lot of people on tumblr.

UPDATE: THIS is (some) of what is going on with tumblr's community labelling (15/04/25)

i meant to update this sooner but i have more information about what is going on now, which i feel is important since more people are actually noticing all the content labelling now and no one seems to have any idea what's going on (and yes, i'm aware probably 2 people will see this but i care too much not to write about it. this post from wip details that the labelling is not from tumblr's policy, but instead from the google play store/android policy- hence the inconsistency between app and browser displayed above. they say, In order for us to remain in Google’s Play Store and for our Tumblr Android app to be available, we needed to make changes that would help us be more compliant with Google’s policies around sensitive content.

i can somewhat understand this? especially since, as i mentioned briefly previously, they want the app to be more appealling to advertisers, ESPECIALLY following recent staff layoffs. losing the android app would presumably lose tumblr a sizeable portion of its userbase which it just can't afford to happen. however, this still does not explain the seeming randomness of the content labelling. i've seen a lot of people concerned that this is another attack on trans people, or on women, and while i don't want to discredit this completely as a theory, i don't think that that's the case (this time), based on the sheer extent and lack of pattern in posts that have been community labelled.[1] my personal theoretical speculation (and please be very aware that this is a speculation rather than a clear conclusive theory with fully supported evidence; do not take my word as fact) is that the community labelling is based on some sort of AI model that's just arbitrarily deciding what posts are "potentially mature" based on patterns that aren't necessarily actually there. hopefully this isn't the case, but i certainly wouldn't be surprised if it is? i don't see how any kind of manual content labelling would have SUCH inconsistencies.

even with this somewhat explained, i think its very right to be concerned about arbitrary content labelling. it doesn't protect anyone. someone who wants to have sensitive/mature content hidden from their dashboard is faced with either having the posts hidden and therefore missing large portions of posts from blogs they follow, or even not being able to see some people's blogs at all, or instead not hiding mature labelled posts and therefore not being able to avoid content they don't want to see. because its not even that all actually nsfw posts are actually being censored, and you can argue whatever you want about whether they should be or not (this is not the point in this specific case) but either way if the point was to follow policy to actually protect users they would be, surely. and yet, here we are.