Tumblr content filtering systems have falsely pointed out publications like “Mature”, users blame the AI

Tumblr is the last technological company to struggle with automated dragges and disassembly that have become unlocked and have raised anger of the users.

In recent days, Tumblr users have complained that their content is reported as “mature”, even if it is not the case. The problem has reduced the visibility of user publications because many people on the platform have configured their parameters to hide the default mature content.

According to many messages from tumblr users impacted, the messages have been falsely reported although they are not sexual or violent, and these false withdrawals have included everything Chat gifs to fandom content has art And even a hand image. Some suspect that AI automation may be to blame for problems.

Similar accusations have tormented other social media platforms in recent weeks. For example, Pinterest finally admitted that an internal error leads to his mass prohibitions. Meanwhile, Instagram this week refused to Comment on your own problems with mass prohibitions What users have said received little coverage or attention outside online complaints on applications like X and Reddit.

In both cases, users suspected that AI moderation was to blame, although Pinterest denied what it is.

On Tumblr, the signaling problem is linked to an update of the Android application, where the company said that it experienced improvements in its mature content filter systems. More specifically, he was testing a new layer of moderation Content labelsTechcrunch told the company.

A Tomblr spokesperson said that the experiences continued and, based on user comments, will be improved before changes move to other platforms.

“While we work to make Tumblr a safer place for all users, we aim to respect a diversified set of interest and content preferences, which can be adjusted in the parameters. We consider this as a continuous process when we continue to refine the way we detect and approach mature content,” said a tumblr spokesperson in an electronic mail.

The company too Recognized the problem on his blogAnswering the question of a user on messages wrongly. Here, the Tumblr team noted that they were aware of “incorrect classification problems” and actively worked to reduce them.

In addition, the post explained that the Tumblr call process was updated in the coming weeks to be able to manage a higher volume of cases. (Tumblr did not answer our questions about the planned changes in the call process.)

That the AI ​​or any other automation is to blame is not clear, because Tumblr would not speak specifically about the cause of new problems. However, the reduction of blog service staff has probably also played a role.

After its 2019 acquisition By WordPress.com Maker Automattic, Tumblr faced dismissals As its staff has been reassigned to Other projects in the parent company. Last year, Automattic announced that the tumblr backend would also be Moved to WordPress To make management easier and endire its financial losses.