X appears to have blocked searches for Taylor Swift as a reaction to a recent trend of graphic AI fakes of the world-famous recording artist being posted to the site. Right now, if you search “Taylor Swift” or “Taylor Swift AI” on X, formerly Twitter, you may see a “Something went wrong” message. After Platformer’s Casey Newton posted about the apparent block, we tried some other searches. If X is intentionally blocking searches for illicit AI-generated images of Swift, it doesn’t seem to have gotten very far.
That’s because if you run into a blocked search term and decide to put the barest amount of effort into trying something else, you’ll get around it. “Taylor AI Swift” or simply putting quotation marks around her name both returned results as of this writing, for example. Even tacking another word onto one of the seemingly blocked search terms gets results. Also, images still show up under the Media tab, although we saw no explicit images when checking there.
Taylor Swift searches don’t return anything. Unless you change the search just a little bit.
Screenshot: Wes Davis / The Verge
A search for “Taylor Swift AI” on Threads.
Screenshot: Wes Davis / The Verge
Both Threads and Instagram suggest “Taylor Swift AI” if you start typing “Taylor” into their search boxes, but neither appears to show results, instead displaying a message saying that the term “is sometimes associated with activities of dangerous organizations and individuals.”
Reports have said that Swift is considering legal action against the sites hosting the images, which users said they preferred Microsoft Designer to make. Microsoft CEO Satya Nadella called the deepfakes “alarming and terrible” in an interview with NBC Nightly News yesterday, and said he believes AI companies need to “move fast” to get better guardrails in place.
White House press secretary Karine Jean-Pierre weighed in yesterday as well, calling on Congress to create legislation to protect people from deepfake porn.
X replied to our email asking whether Swift’s name is being blocked intentionally with an emailed auto-reply.
Update January 27th, 2024, 3:22PM ET: Added additional context.
