This Tumblr May Contain Sensitive Media

If you were on Tumblr between, say, 2012 and 2018, you know the drill.

Tap to view. Are you sure? Are you really sure? this tumblr may contain sensitive media

We didn’t know it then, but that little warning was a kind of farewell. A reminder that the wild, weird, unregulated internet was already being boxed up — one blurred post at a time. If you were on Tumblr between, say, 2012

So here’s to that goofy gray box. To the art it hid and the communities it hurt. To the bots that flagged a statue’s nipple but not actual harassment. To the dashboard refugees who migrated to Twitter, then Discord, then nowhere at all. Are you really sure

But Tumblr’s version was different. It was clunky. Honest in its clunkiness. It didn’t pretend to be smart. It just asked: Are you over 18? Do you accept the risk?

The primary catalyst for this message is Tumblr’s internal machine-learning classifier. The algorithm scans media attachments—such as image sets, text blocks, and video uploads—unilaterally marking items that resemble sexually explicit material (NSFW), violence, or sensitive themes.