Tumblr’s content filtering systems have been falsely flagging posts as ‘mature,’ users blame AI

Tumblr is the latest tech company to grapple with automated flagging and takedowns that have gone haywire and raised the ire of users.

In recent days, Tumblr users have complained their content is being flagged as “mature,” even when that’s not the case. The problem has reduced the visibility of users’ posts because many people on the platform have configured their settings to hide mature content by default.

According to numerous posts from impacted Tumblr users, posts have been falsely flagged despite not being sexual or violent, and these false takedowns have included everything from cat GIFs to fandom content to art and even a picture of hands. Some suspect that AI-based automation could be to blame for the issues.

Similar accusations have plagued other social media platforms in recent weeks. For instance, Pinterest finally admitted that an internal error led to its mass user bans. Meanwhile, Instagram this week declined to comment on its own trouble with mass bans that users said have been given little coverage or attention outside of online complaints on apps like X and Reddit.

In both cases, users suspected that AI-based moderation was to blame, though Pinterest denied that was the case.

On Tumblr, the flagging issue is tied to an update to the Android app, where the company said it’s been experimenting with improvements to its mature content filtering systems. Specifically, it was testing out a new layer of moderation to Content Labels, the company told TechCrunch.

A spokesperson for Tumblr said the experiments are continuing and, based on user feedback, will be improved before the changes roll out to other platforms.

“As we work to make Tumblr a safer place for all users, we aim to respect a diverse set of interests and content preferences, which can be adjusted in settings. We view this as an ongoing process as we continue to fine-tune how we detect and address mature content,” a Tumblr spokesperson said in an emailed statement.

The company also acknowledged the problem on its blog, responding to a user’s question about the falsely flagged posts. Here, Tumblr’s team noted they were aware of the “incorrect classification issues” and were actively working to reduce them.

In addition, the post explained that Tumblr’s appeal process was being updated in the weeks ahead to be able to handle a higher volume of cases. (Tumblr did not respond to our questions about the appeals process’ planned changes.)

Whether AI or other automation is to blame is not clear, as Tumblr wouldn’t speak specifically to the cause of the new issues. However, the reduced staffing at the blogging service has likely also played a role.

Following its 2019 acquisition by WordPress.com maker Automattic, Tumblr has faced layoffs as its staff was reassigned to other projects at the parent company. Last year, Automattic announced that Tumblr’s backend would also be moved to WordPress to make management easier and stem its financial losses.