The latest tech company to struggle with automated flagging and takedowns is Tumblr, which has sparked outrage among its users due to the errors and inconsistencies in the process.

Recently, numerous Tumblr users have reported that their content has been mistakenly labeled as “mature,” resulting in reduced visibility of their posts, as many users have set their default settings to hide mature content.

According to multiple posts from affected users, a wide range of content has been incorrectly flagged, including cat GIFs, fan-created content, artwork, and even a picture of hands. Some users suspect that AI-driven automation may be the root cause of the issue.

Similar issues have been reported on other social media platforms in recent weeks. For example, Pinterest recently acknowledged that an internal error led to mass user bans, while Instagram has declined to comment on its own issues with mass bans that users claim have received little attention outside of online complaints.

In both cases, users have speculated that AI-based moderation may be to blame, although Pinterest has denied this.

On Tumblr, the flagging issue is linked to an update to the Android app, where the company has been testing improvements to its mature content filtering systems. Specifically, the company has been experimenting with a new layer of moderation for Content Labels, according to a statement from Tumblr.

A Tumblr spokesperson stated that the experiments are ongoing and will be refined based on user feedback before being rolled out to other platforms.

“As we strive to make Tumblr a safer environment for all users, we aim to respect a diverse range 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 addressed the issue on its blog, responding to a user’s question about the falsely flagged posts. In the post, Tumblr’s team acknowledged the “incorrect classification issues” and stated that they are actively working to reduce them.

Additionally, the post explained that Tumblr’s appeal process will be updated in the coming weeks to handle a higher volume of cases. (Tumblr did not respond to our questions about the planned changes to the appeals process.)

While it is unclear whether AI or other automation is the cause of the issue, as Tumblr declined to comment on the specific cause, the reduction in staffing at the blogging service has likely contributed to the problem.

Following its acquisition by Automattic in 2019, Tumblr has experienced layoffs as its staff was reassigned to other projects at the parent company. Last year, Automattic announced that Tumblr’s back-end infrastructure would be migrated to WordPress to simplify management and mitigate financial losses.