Tumblr’s content-filtering methods have been falsely flagging posts as ‘mature,’ customers blame AI
Tumblr is the newest tech firm to grapple with automated flagging and takedowns which have gone haywire and raised the ire of customers.
In latest days, Tumblr customers have complained their content material is being flagged as “mature,” even when that’s not the case. The issue has decreased the visibility of customers’ posts as a result of many individuals on the platform have configured their settings to cover mature content material by default.
In line with quite a few posts from impacted Tumblr customers, posts have been falsely flagged regardless of not being sexual or violent, and these false takedowns have included all the pieces from cat GIFs to fandom content material to artwork and even an image of palms. Some suspect that AI-based automation may very well be responsible for the problems.
Related accusations have plagued different social media platforms in latest weeks. As an illustration, Pinterest lastly admitted that an inner error led to its mass person bans. In the meantime, Instagram this week declined to remark by itself bother with mass bans that customers mentioned have been given little protection or consideration outdoors of on-line complaints on apps like X and Reddit.
In each instances, customers suspected that AI-based moderation was responsible, although Pinterest denied that was the case.
On Tumblr, the flagging challenge is tied to an replace to the Android app, the place the corporate mentioned it’s been experimenting with enhancements to its mature content material filtering methods. Particularly, it was testing out a brand new layer of moderation to Content material Labels, the corporate advised TechCrunch.
A spokesperson for Tumblr mentioned the experiments are persevering with and, primarily based on person suggestions, might be improved earlier than the modifications roll out to different platforms.
“As we work to make Tumblr a safer place for all customers, we purpose to respect a various set of pursuits and content material preferences, which might be adjusted in settings. We view this as an ongoing course of as we proceed to fine-tune how we detect and deal with mature content material,” a Tumblr spokesperson mentioned in an emailed assertion.
The corporate additionally acknowledged the issue on its weblog, responding to a person’s query concerning the falsely flagged posts. Right here, Tumblr’s group famous they have been conscious of the “incorrect classification points” and have been actively working to cut back them.
As well as, the put up defined that Tumblr’s enchantment course of was being up to date within the weeks forward to have the ability to deal with a better quantity of instances. (Tumblr didn’t reply to our questions concerning the appeals course of’ deliberate modifications.)
Whether or not AI or different automation is responsible is just not clear, as Tumblr wouldn’t converse particularly to the reason for the brand new points. Nevertheless, the decreased staffing on the running a blog service has probably additionally performed a job.
Following its 2019 acquisition by WordPress.com maker Automattic, Tumblr has confronted layoffs as its workers was reassigned to different initiatives on the dad or mum firm. Final yr, Automattic introduced that Tumblr’s again finish would even be moved to WordPress to make administration simpler and stem its monetary losses.