Archive — US — techcrunch
Tumblr’s content-filtering systems have been falsely flagging posts as ‘mature,’users blame AI
Tumblr users are frustrated as the platform's content-filtering system erroneously flags numerous non-mature posts as 'mature', causing reduced visibility. The issue, linked to recent updates in their Android app, has raised suspicions of AI errors in moderation, similar to issues faced by other social media platforms. Tumblr acknowledges the problem and is working on improvements and an updated appeals process.
Key points
- Many Tumblr users report that their posts are incorrectly labeled as 'mature', which affects visibility.
- The incorrect flagging is impacting a variety of content including GIFs, fandom items, and art.
- AI-based moderation is suspected to be causing the issues across multiple social media platforms.
- Tumblr attributes the problem to updates in their Android app aimed at improving content filtering.
- The company plans to enhance the filtering system based on user feedback and update their appeals process to handle more cases.
- Staff reductions since Tumblr's acquisition by Automattic might also be affecting their moderation capabilities.
Why it matters
- The incident reflects a growing trend of automated moderation failures in social media, prompting discussions about AI reliability.
- Pinterest recently acknowledged user bans due to an internal error, drawing parallels to Tumblr's situation.
- Social media platforms are facing increasing scrutiny for their content moderation practices, especially regarding AI use and user frustrations.
Tumblr's false flagging issue highlights challenges in automated moderation across social media and the need for better AI oversight and user support, as the company works to refine its systems amidst reduced staffing.